From 41091087396f935d4acf70b018ba54889fcf55a1 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Fri, 12 Jun 2009 19:08:44 +0200
Subject: r300: add support for EXT_texture_sRGB

Tested with glean/texture_srgb and wine/d3d9 tests on RV535
---
 src/mesa/drivers/dri/r300/r300_context.c  | 1 +
 src/mesa/drivers/dri/r300/r300_reg.h      | 2 ++
 src/mesa/drivers/dri/r300/r300_texstate.c | 4 ++++
 3 files changed, 7 insertions(+)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index 394521a051..76881e4928 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -124,6 +124,7 @@ const struct dri_extension card_extensions[] = {
   {"GL_EXT_texture_lod_bias",		NULL},
   {"GL_EXT_texture_mirror_clamp",	NULL},
   {"GL_EXT_texture_rectangle",		NULL},
+  {"GL_EXT_texture_sRGB",		NULL},
   {"GL_EXT_vertex_array_bgra",		NULL},
   {"GL_ATI_separate_stencil",		GL_ATI_separate_stencil_functions},
   {"GL_ATI_texture_env_combine3",	NULL},
diff --git a/src/mesa/drivers/dri/r300/r300_reg.h b/src/mesa/drivers/dri/r300/r300_reg.h
index c22616b95f..357c600af9 100644
--- a/src/mesa/drivers/dri/r300/r300_reg.h
+++ b/src/mesa/drivers/dri/r300/r300_reg.h
@@ -1467,6 +1467,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R300_TX_FORMAT_3D		   (1 << 25)
 #	define R300_TX_FORMAT_CUBIC_MAP		   (2 << 25)
 
+#	define R300_TX_FORMAT_GAMMA			(1 << 21)
+
 	/* gap */
 	/* Floating point formats */
 	/* Note - hardware supports both 16 and 32 bit floating point */
diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c
index 6d6a90aa88..6e47321246 100644
--- a/src/mesa/drivers/dri/r300/r300_texstate.c
+++ b/src/mesa/drivers/dri/r300/r300_texstate.c
@@ -119,6 +119,10 @@ static const struct tx_table {
 	_ASSIGN(Z24_S8, R300_EASY_TX_FORMAT(X, X, X, X, X24_Y8)),
 	_ASSIGN(S8_Z24, R300_EASY_TX_FORMAT(Y, Y, Y, Y, X24_Y8)),
 	_ASSIGN(Z32, R300_EASY_TX_FORMAT(X, X, X, X, X32)),
+	/* EXT_texture_sRGB */
+	_ASSIGN(SRGBA8, R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8) | R300_TX_FORMAT_GAMMA),
+	_ASSIGN(SLA8, R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8) | R300_TX_FORMAT_GAMMA),
+	_ASSIGN(SL8, R300_EASY_TX_FORMAT(X, X, X, ONE, X8) | R300_TX_FORMAT_GAMMA),
 	/* *INDENT-ON* */
 };
 
-- 
cgit v1.2.3


From 2506c4e8b142b933668db2b478333ebdfcfd0d96 Mon Sep 17 00:00:00 2001
From: Jerome Glisse <jglisse@redhat.com>
Date: Wed, 17 Jun 2009 13:51:33 +0200
Subject: r300: don't emit vap index offset on r5xx hw when using cs

vap index offset is programmed to 0 by the kernel, it
would add work to kernel checker to allow userspace
programming of this so it's now disallowed with CS
on KMS.
---
 src/mesa/drivers/dri/r300/r300_cmdbuf.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
index 0261a5b1d8..b5c6bd1835 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
@@ -503,7 +503,7 @@ void r300InitCmdBuf(r300ContextPtr r300)
 	r300->hw.vap_cntl.cmd[R300_VAP_CNTL_FLUSH] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_PVS_STATE_FLUSH_REG, 1);
 	r300->hw.vap_cntl.cmd[R300_VAP_CNTL_FLUSH_1] = 0;
 	r300->hw.vap_cntl.cmd[R300_VAP_CNTL_CMD] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_CNTL, 1);
-	if (is_r500) {
+	if (is_r500 && !r300->radeon.radeonScreen->kernel_mm) {
 	    ALLOC_STATE(vap_index_offset, always, 2, 0);
 	    r300->hw.vap_index_offset.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R500_VAP_INDEX_OFFSET, 1);
 	    r300->hw.vap_index_offset.cmd[1] = 0;
-- 
cgit v1.2.3


From 46000cecc32104702fcb0de5a842d11b18c41c6b Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Thu, 18 Jun 2009 13:20:32 +1000
Subject: r300: use vbo_split_prims to split up large vertex buffers.

This lets ut2004 avoid hitting the elt warning.
---
 src/mesa/drivers/dri/r300/r300_draw.c   | 9 +++++++++
 src/mesa/drivers/dri/r300/r300_render.c | 4 ++--
 2 files changed, 11 insertions(+), 2 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c
index cc5650fb7c..92bb0ee338 100644
--- a/src/mesa/drivers/dri/r300/r300_draw.c
+++ b/src/mesa/drivers/dri/r300/r300_draw.c
@@ -452,12 +452,21 @@ static void r300DrawPrims(GLcontext *ctx,
 			 GLuint min_index,
 			 GLuint max_index)
 {
+	struct split_limits limits;
 	GLboolean retval;
 
+	limits.max_verts = 65535;
+	limits.max_indices = 65535;
+	limits.max_vb_size = 1024*1024;
+
 	if (min_index) {
 		vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, min_index, max_index, r300DrawPrims );
 		return;
 	}
+	if ((ib && ib->count > 65536)) {
+		vbo_split_prims (ctx, arrays, prim, nr_prims, ib, min_index, max_index, r300DrawPrims, &limits);
+		return;
+	}
 
 	/* Make an attempt at drawing */
 	retval = r300TryDrawPrims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index 1356305a21..bf50b062f6 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -243,7 +243,6 @@ static void r300EmitAOS(r300ContextPtr rmesa, GLuint nr, GLuint offset)
 		fprintf(stderr, "%s: nr=%d, ofs=0x%08x\n", __FUNCTION__, nr,
 			offset);
 
-
 	if (!rmesa->radeon.radeonScreen->kernel_mm) {
 		BEGIN_BATCH(sz+2+(nr * 2));
 		OUT_BATCH_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, sz - 1);
@@ -384,7 +383,8 @@ void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim)
 		 * arrays. *sigh*
 		 */
 		r300EmitElts(ctx, num_verts);
-		r300EmitAOS(rmesa, rmesa->radeon.tcl.aos_count, start);
+		/* don't pass start if we are split up */
+		r300EmitAOS(rmesa, rmesa->radeon.tcl.aos_count, 0);
 		if (rmesa->radeon.radeonScreen->kernel_mm) {
 			BEGIN_BATCH_NO_AUTOSTATE(2);
 			OUT_BATCH_REGSEQ(R300_VAP_VF_MAX_VTX_INDX, 1);
-- 
cgit v1.2.3


From 43bb78f2bb6c851d989903e7eb996e87113d878c Mon Sep 17 00:00:00 2001
From: Roland Scheidegger <sroland@vmware.com>
Date: Fri, 19 Jun 2009 20:00:55 +0200
Subject: radeons: use dp4 for position invariant vertex programs

Fixes #22181. R200 requires this since DP4 is used in hw tnl mode.
R300 prefers it (should be faster due to no instruction dependencies), but
both methods should be correct (when sw tcl is used though, MUL/MAD might
be faster). Probably doesn't make much difference for R100 since vertex progs
are executed in software anyway, but let's just keep it the same there too.
---
 src/mesa/drivers/dri/r200/r200_context.c     | 2 ++
 src/mesa/drivers/dri/r300/r300_context.c     | 2 ++
 src/mesa/drivers/dri/radeon/radeon_context.c | 2 ++
 3 files changed, 6 insertions(+)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c
index c06751516e..058296eab2 100644
--- a/src/mesa/drivers/dri/r200/r200_context.c
+++ b/src/mesa/drivers/dri/r200/r200_context.c
@@ -405,6 +405,8 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
 
    ctx->Const.MaxDrawBuffers = 1;
 
+   _mesa_set_mvp_with_dp4( ctx, GL_TRUE );
+
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index 8f0effd83e..7d6705058f 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -327,6 +327,8 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 
 	ctx->Const.MaxDrawBuffers = 1;
 
+	_mesa_set_mvp_with_dp4( ctx, GL_TRUE );
+
 	/* Initialize the software rasterizer and helper modules.
 	 */
 	_swrast_CreateContext(ctx);
diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c
index ea81a3250b..b67bda7d06 100644
--- a/src/mesa/drivers/dri/radeon/radeon_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_context.c
@@ -355,6 +355,8 @@ radeonCreateContext( const __GLcontextModes *glVisual,
 
    ctx->Const.MaxDrawBuffers = 1;
 
+   _mesa_set_mvp_with_dp4( ctx, GL_TRUE );
+
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-- 
cgit v1.2.3


From 7ce814b25f8c216c7897904cbce7f570112e60ef Mon Sep 17 00:00:00 2001
From: Roland Scheidegger <sroland@vmware.com>
Date: Fri, 19 Jun 2009 20:00:55 +0200
Subject: radeons: use dp4 for position invariant vertex programs

Fixes #22181. R200 requires this since DP4 is used in hw tnl mode.
R300 prefers it (should be faster due to no instruction dependencies), but
both methods should be correct (when sw tcl is used though, MUL/MAD might
be faster). Probably doesn't make much difference for R100 since vertex progs
are executed in software anyway, but let's just keep it the same there too.
---
 src/mesa/drivers/dri/r200/r200_context.c     | 2 ++
 src/mesa/drivers/dri/r300/r300_context.c     | 2 ++
 src/mesa/drivers/dri/radeon/radeon_context.c | 2 ++
 3 files changed, 6 insertions(+)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c
index 8924849d08..241390cef7 100644
--- a/src/mesa/drivers/dri/r200/r200_context.c
+++ b/src/mesa/drivers/dri/r200/r200_context.c
@@ -388,6 +388,8 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
 
    ctx->Const.MaxDrawBuffers = 1;
 
+   _mesa_set_mvp_with_dp4( ctx, GL_TRUE );
+
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index 76881e4928..7aa8075aa7 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -410,6 +410,8 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 
 	r300InitConstValues(ctx, screen);
 
+	_mesa_set_mvp_with_dp4( ctx, GL_TRUE );
+
 	/* Initialize the software rasterizer and helper modules.
 	 */
 	_swrast_CreateContext(ctx);
diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c
index 8f780c443c..229b4387ca 100644
--- a/src/mesa/drivers/dri/radeon/radeon_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_context.c
@@ -316,6 +316,8 @@ r100CreateContext( const __GLcontextModes *glVisual,
 
    ctx->Const.MaxDrawBuffers = 1;
 
+   _mesa_set_mvp_with_dp4( ctx, GL_TRUE );
+
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-- 
cgit v1.2.3


From 43b3b745e4d893b64d6331cb6183f8615e613f66 Mon Sep 17 00:00:00 2001
From: Roland Scheidegger <sroland@vmware.com>
Date: Thu, 25 Jun 2009 15:57:33 +0200
Subject: radeon: fix hw texture limits

still always enable max, but the right values this time.
More work should probably be done for saner limits without mm, and/or
dri conf option allow_large_textures (which is ignored) removed.
3D limit on r100 is pretty arbitrary as still handled by swrast anyway.
Also fix r300 limits (except 3d I've no idea what the max is anyway so
keep using mesa default).
---
 src/mesa/drivers/dri/r200/r200_context.c     |  6 +++---
 src/mesa/drivers/dri/r300/r300_context.c     | 10 ++++++++--
 src/mesa/drivers/dri/radeon/radeon_context.c |  7 +++----
 3 files changed, 14 insertions(+), 9 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c
index 1ba8e5e612..9a92a32079 100644
--- a/src/mesa/drivers/dri/r200/r200_context.c
+++ b/src/mesa/drivers/dri/r200/r200_context.c
@@ -356,9 +356,9 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
 
    /* FIXME: When no memory manager is available we should set this 
     * to some reasonable value based on texture memory pool size */
-   ctx->Const.MaxTextureLevels = 11;
-   ctx->Const.Max3DTextureLevels = 8;
-   ctx->Const.MaxCubeTextureLevels = 11;
+   ctx->Const.MaxTextureLevels = 12;
+   ctx->Const.Max3DTextureLevels = 9;
+   ctx->Const.MaxCubeTextureLevels = 12;
    ctx->Const.MaxTextureRectSize = 2048;
 
    ctx->Const.MaxTextureMaxAnisotropy = 16.0;
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index 14a11ea1fb..b7d75426c5 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -270,10 +270,16 @@ static void r300InitConstValues(GLcontext *ctx, radeonScreenPtr screen)
 	ctx->Const.MaxTextureMaxAnisotropy = 16.0;
 	ctx->Const.MaxTextureLodBias = 16.0;
 
-	if (screen->chip_family >= CHIP_FAMILY_RV515)
+	if (screen->chip_family >= CHIP_FAMILY_RV515) {
 		ctx->Const.MaxTextureLevels = 13;
-	else
+		ctx->Const.MaxCubeTextureLevels = 13;
+		ctx->Const.MaxTextureRectSize = 4096;
+	}
+	else {
 		ctx->Const.MaxTextureLevels = 12;
+		ctx->Const.MaxCubeTextureLevels = 12;
+		ctx->Const.MaxTextureRectSize = 2048;
+	}
 
 	ctx->Const.MinPointSize = 1.0;
 	ctx->Const.MinPointSizeAA = 1.0;
diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c
index 3d03f6266d..46cba73e29 100644
--- a/src/mesa/drivers/dri/radeon/radeon_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_context.c
@@ -285,10 +285,9 @@ r100CreateContext( const __GLcontextModes *glVisual,
 
    /* FIXME: When no memory manager is available we should set this 
     * to some reasonable value based on texture memory pool size */
-   /* FIXME: does r100 support 2048x2048 texture ? */
-   ctx->Const.MaxTextureLevels = 11;
-   ctx->Const.Max3DTextureLevels = 8;
-   ctx->Const.MaxCubeTextureLevels = 11;
+   ctx->Const.MaxTextureLevels = 12;
+   ctx->Const.Max3DTextureLevels = 9;
+   ctx->Const.MaxCubeTextureLevels = 12;
    ctx->Const.MaxTextureRectSize = 2048;
 
    ctx->Const.MaxTextureMaxAnisotropy = 16.0;
-- 
cgit v1.2.3


From 9fa0d25c547a940fa275f786a63de85f6bc39870 Mon Sep 17 00:00:00 2001
From: Nicolai Hähnle <nhaehnle@gmail.com>
Date: Sat, 27 Jun 2009 18:08:44 +0200
Subject: radeon: Update .gitignore
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Add all source files that are symlink'ed from common radeon code to the
ignore list.

Signed-off-by: Nicolai Hähnle <nhaehnle@gmail.com>
---
 src/mesa/drivers/dri/r200/.gitignore | 14 +++++++++++++-
 src/mesa/drivers/dri/r300/.gitignore | 13 ++++++++++++-
 2 files changed, 25 insertions(+), 2 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r200/.gitignore b/src/mesa/drivers/dri/r200/.gitignore
index 3773d8ea73..2f9cd1a987 100644
--- a/src/mesa/drivers/dri/r200/.gitignore
+++ b/src/mesa/drivers/dri/r200/.gitignore
@@ -1,3 +1,15 @@
+radeon_bocs_wrapper.h
+radeon_bo_legacy.[ch]
 radeon_chipset.h
-radeon_screen.*
+radeon_cmdbuf.h
+radeon_common.[ch]
+radeon_common_context.[ch]
+radeon_cs_legacy.[ch]
+radeon_dma.[ch]
+radeon_fbo.c
+radeon_lock.[ch]
+radeon_mipmap_tree.[ch]
+radeon_screen.[ch]
+radeon_span.[ch]
+radeon_texture.[ch]
 server
diff --git a/src/mesa/drivers/dri/r300/.gitignore b/src/mesa/drivers/dri/r300/.gitignore
index 3689a6a78e..2f9cd1a987 100644
--- a/src/mesa/drivers/dri/r300/.gitignore
+++ b/src/mesa/drivers/dri/r300/.gitignore
@@ -1,4 +1,15 @@
+radeon_bocs_wrapper.h
+radeon_bo_legacy.[ch]
 radeon_chipset.h
+radeon_cmdbuf.h
+radeon_common.[ch]
+radeon_common_context.[ch]
+radeon_cs_legacy.[ch]
+radeon_dma.[ch]
+radeon_fbo.c
+radeon_lock.[ch]
+radeon_mipmap_tree.[ch]
 radeon_screen.[ch]
-radeon_span.h
+radeon_span.[ch]
+radeon_texture.[ch]
 server
-- 
cgit v1.2.3


From 2ed3eddf9a828f2ff6c74b0913ca37fb60672950 Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Thu, 2 Jul 2009 20:44:30 +1000
Subject: radeon/r300: use base width/height.

I suspect this might break TFP in some way but it makes firecube run here
---
 src/mesa/drivers/dri/r300/r300_cmdbuf.c         | 21 +++++++++++----------
 src/mesa/drivers/dri/radeon/radeon_state_init.c |  4 ++--
 2 files changed, 13 insertions(+), 12 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
index b5c6bd1835..19d240f576 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
@@ -232,8 +232,8 @@ void r300_emit_scissor(GLcontext *ctx)
     } else {
         x1 = 0;
         y1 = 0;
-        x2 = rrb->width - 1;
-        y2 = rrb->height - 1;
+        x2 = rrb->base.Width - 1;
+        y2 = rrb->base.Height - 1;
     }
     if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) {
         x1 += R300_SCISSORS_OFFSET;
@@ -263,7 +263,8 @@ static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
 		fprintf(stderr, "no rrb\n");
 		return;
 	}
-
+	
+	fprintf(stderr,"rrb is %p %d %dx%d\n", rrb, offset, rrb->base.Width, rrb->base.Height);
 	cbpitch = (rrb->pitch / rrb->cpp);
 	if (rrb->cpp == 4)
 		cbpitch |= R300_COLOR_FORMAT_ARGB8888;
@@ -289,14 +290,14 @@ static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
             BEGIN_BATCH_NO_AUTOSTATE(3);
             OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2);
             OUT_BATCH(0);
-            OUT_BATCH(((rrb->width - 1) << R300_SCISSORS_X_SHIFT) |
-                    ((rrb->height - 1) << R300_SCISSORS_Y_SHIFT));
+            OUT_BATCH(((rrb->base.Width - 1) << R300_SCISSORS_X_SHIFT) |
+                    ((rrb->base.Height - 1) << R300_SCISSORS_Y_SHIFT));
             END_BATCH();
             BEGIN_BATCH_NO_AUTOSTATE(16);
             for (i = 0; i < 4; i++) {
                 OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2);
                 OUT_BATCH((0 << R300_CLIPRECT_X_SHIFT) | (0 << R300_CLIPRECT_Y_SHIFT));
-                OUT_BATCH(((rrb->width - 1) << R300_CLIPRECT_X_SHIFT) | ((rrb->height - 1) << R300_CLIPRECT_Y_SHIFT));
+                OUT_BATCH(((rrb->base.Width - 1) << R300_CLIPRECT_X_SHIFT) | ((rrb->base.Height - 1) << R300_CLIPRECT_Y_SHIFT));
             }
             OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1);
             OUT_BATCH(0xAAAA);
@@ -308,15 +309,15 @@ static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
             OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2);
             OUT_BATCH((R300_SCISSORS_OFFSET << R300_SCISSORS_X_SHIFT) |
                     (R300_SCISSORS_OFFSET << R300_SCISSORS_Y_SHIFT));
-            OUT_BATCH(((rrb->width + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_X_SHIFT) |
-                    ((rrb->height + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_Y_SHIFT));
+            OUT_BATCH(((rrb->base.Width + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_X_SHIFT) |
+                    ((rrb->base.Height + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_Y_SHIFT));
             END_BATCH();
             BEGIN_BATCH_NO_AUTOSTATE(16);
             for (i = 0; i < 4; i++) {
                 OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2);
                 OUT_BATCH((R300_SCISSORS_OFFSET << R300_CLIPRECT_X_SHIFT) | (R300_SCISSORS_OFFSET << R300_CLIPRECT_Y_SHIFT));
-                OUT_BATCH(((R300_SCISSORS_OFFSET + rrb->width - 1) << R300_CLIPRECT_X_SHIFT) |
-                          ((R300_SCISSORS_OFFSET + rrb->height - 1) << R300_CLIPRECT_Y_SHIFT));
+                OUT_BATCH(((R300_SCISSORS_OFFSET + rrb->base.Width - 1) << R300_CLIPRECT_X_SHIFT) |
+                          ((R300_SCISSORS_OFFSET + rrb->base.Height - 1) << R300_CLIPRECT_Y_SHIFT));
             }
             OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1);
             OUT_BATCH(0xAAAA);
diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c
index c517487098..a55cd6dfcb 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state_init.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c
@@ -453,8 +453,8 @@ static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
    OUT_BATCH(0);
    OUT_BATCH(CP_PACKET0(RADEON_RE_WIDTH_HEIGHT, 0));
    if (rrb) {
-       OUT_BATCH(((rrb->width - 1) << RADEON_RE_WIDTH_SHIFT) |
-                 ((rrb->height - 1) << RADEON_RE_HEIGHT_SHIFT));
+       OUT_BATCH(((rrb->base.Width - 1) << RADEON_RE_WIDTH_SHIFT) |
+                 ((rrb->base.Height - 1) << RADEON_RE_HEIGHT_SHIFT));
    } else {
        OUT_BATCH(0);
    }
-- 
cgit v1.2.3


From 54ee188a00e31d239cbd256e7ba5ffd2c1259650 Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Thu, 2 Jul 2009 20:57:45 +1000
Subject: radeon/r200/r300: drop radeon renderbuffer private width/height

half stealing the code without taking the intel regions
---
 src/mesa/drivers/dri/r200/r200_texstate.c           |  6 +++---
 src/mesa/drivers/dri/r300/r300_texstate.c           | 10 +++++-----
 src/mesa/drivers/dri/radeon/radeon_common_context.c |  4 ++--
 src/mesa/drivers/dri/radeon/radeon_common_context.h |  2 --
 src/mesa/drivers/dri/radeon/radeon_texstate.c       |  6 +++---
 5 files changed, 13 insertions(+), 15 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r200/r200_texstate.c b/src/mesa/drivers/dri/r200/r200_texstate.c
index ed1995e147..4e53672aee 100644
--- a/src/mesa/drivers/dri/r200/r200_texstate.c
+++ b/src/mesa/drivers/dri/r200/r200_texstate.c
@@ -834,7 +834,7 @@ void r200SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo
 		rImage->mt = NULL;
 	}
 	_mesa_init_teximage_fields(radeon->glCtx, target, texImage,
-				   rb->width, rb->height, 1, 0, rb->cpp);
+				   rb->base.Width, rb->base.Height, 1, 0, rb->cpp);
 	texImage->RowStride = rb->pitch / rb->cpp;
 	texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx,
 							internalFormat,
@@ -866,8 +866,8 @@ void r200SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo
 		t->pp_txfilter |= tx_table_le[MESA_FORMAT_RGB565].filter;
 		break;
 	}
-        t->pp_txsize = ((rb->width - 1) << RADEON_TEX_USIZE_SHIFT)
-		   | ((rb->height - 1) << RADEON_TEX_VSIZE_SHIFT);
+        t->pp_txsize = ((rb->base.Width - 1) << RADEON_TEX_USIZE_SHIFT)
+		   | ((rb->base.Height - 1) << RADEON_TEX_VSIZE_SHIFT);
         t->pp_txformat |= R200_TXFORMAT_NON_POWER2;
 	t->pp_txpitch = pitch_val;
         t->pp_txpitch -= 32;
diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c
index 6e47321246..c380840a00 100644
--- a/src/mesa/drivers/dri/r300/r300_texstate.c
+++ b/src/mesa/drivers/dri/r300/r300_texstate.c
@@ -437,7 +437,7 @@ void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo
 		rImage->mt = NULL;
 	}
 	_mesa_init_teximage_fields(radeon->glCtx, target, texImage,
-				   rb->width, rb->height, 1, 0, rb->cpp);
+				   rb->base.Width, rb->base.Height, 1, 0, rb->cpp);
 	texImage->RowStride = rb->pitch / rb->cpp;
 	texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx,
 							internalFormat,
@@ -473,15 +473,15 @@ void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo
 		break;
 	}
 	pitch_val--;
-	t->pp_txsize = ((rb->width - 1) << R300_TX_WIDTHMASK_SHIFT) |
-              ((rb->height - 1) << R300_TX_HEIGHTMASK_SHIFT);
+	t->pp_txsize = ((rb->base.Width - 1) << R300_TX_WIDTHMASK_SHIFT) |
+              ((rb->base.Height - 1) << R300_TX_HEIGHTMASK_SHIFT);
 	t->pp_txsize |= R300_TX_SIZE_TXPITCH_EN;
 	t->pp_txpitch |= pitch_val;
 
 	if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
-	    if (rb->width > 2048)
+	    if (rb->base.Width > 2048)
 		t->pp_txpitch |= R500_TXWIDTH_BIT11;
-	    if (rb->height > 2048)
+	    if (rb->base.Height > 2048)
 		t->pp_txpitch |= R500_TXHEIGHT_BIT11;
 	}
 	t->validated = GL_TRUE;
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c
index 009859feca..94bda78ce3 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c
@@ -605,8 +605,8 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 
 		rb->cpp = buffers[i].cpp;
 		rb->pitch = buffers[i].pitch;
-		rb->width = drawable->w;
-		rb->height = drawable->h;
+		rb->base.Width = drawable->w;
+		rb->base.Height = drawable->h;
 		rb->has_surface = 0;
 
 		if (buffers[i].attachment == __DRI_BUFFER_STENCIL && depth_bo) {
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h
index 061168fe96..b607fad87b 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common_context.h
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h
@@ -83,8 +83,6 @@ struct radeon_renderbuffer
 	unsigned int cpp;
 	/* unsigned int offset; */
 	unsigned int pitch;
-	unsigned int width;
-	unsigned int height;
 
 	uint32_t draw_offset; /* FBO */
 	/* boo Xorg 6.8.2 compat */
diff --git a/src/mesa/drivers/dri/radeon/radeon_texstate.c b/src/mesa/drivers/dri/radeon/radeon_texstate.c
index 279bcd4df6..d33eb9988f 100644
--- a/src/mesa/drivers/dri/radeon/radeon_texstate.c
+++ b/src/mesa/drivers/dri/radeon/radeon_texstate.c
@@ -706,7 +706,7 @@ void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_
 		rImage->mt = NULL;
 	}
 	_mesa_init_teximage_fields(radeon->glCtx, target, texImage,
-				   rb->width, rb->height, 1, 0, rb->cpp);
+				   rb->base.Width, rb->base.Height, 1, 0, rb->cpp);
 	texImage->RowStride = rb->pitch / rb->cpp;
 	texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx,
 							internalFormat,
@@ -738,8 +738,8 @@ void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_
 		t->pp_txfilter |= tx_table[MESA_FORMAT_RGB565].filter;
 		break;
 	}
-        t->pp_txsize = ((rb->width - 1) << RADEON_TEX_USIZE_SHIFT)
-		   | ((rb->height - 1) << RADEON_TEX_VSIZE_SHIFT);
+        t->pp_txsize = ((rb->base.Width - 1) << RADEON_TEX_USIZE_SHIFT)
+		   | ((rb->base.Height - 1) << RADEON_TEX_VSIZE_SHIFT);
         t->pp_txformat |= RADEON_TXFORMAT_NON_POWER2;
 	t->pp_txpitch = pitch_val;
         t->pp_txpitch -= 32;
-- 
cgit v1.2.3


From 6b2461fec905b7dc25ad051d602b440503f544ea Mon Sep 17 00:00:00 2001
From: Michel Dänzer <daenzer@vmware.com>
Date: Fri, 3 Jul 2009 12:44:02 +0200
Subject: r300: Guard debugging output.

---
 src/mesa/drivers/dri/r300/r300_cmdbuf.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
index 19d240f576..90ca45b3e7 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
@@ -263,8 +263,9 @@ static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
 		fprintf(stderr, "no rrb\n");
 		return;
 	}
-	
-	fprintf(stderr,"rrb is %p %d %dx%d\n", rrb, offset, rrb->base.Width, rrb->base.Height);
+
+        if (RADEON_DEBUG & DEBUG_STATE)
+           fprintf(stderr,"rrb is %p %d %dx%d\n", rrb, offset, rrb->base.Width, rrb->base.Height);
 	cbpitch = (rrb->pitch / rrb->cpp);
 	if (rrb->cpp == 4)
 		cbpitch |= R300_COLOR_FORMAT_ARGB8888;
-- 
cgit v1.2.3


From 6fff62ee3fdbfe7d8ba15d3ad001f9afd120c307 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Mon, 29 Jun 2009 19:50:39 +0200
Subject: r300: fix vertex limits

- don't limit vertex count if we are using indices
- max indices count is 65535 not 65536
- remove some comments that don't apply anymore
- remove unreachable code
---
 src/mesa/drivers/dri/r300/r300_draw.c   | 10 ++++++----
 src/mesa/drivers/dri/r300/r300_render.c | 19 +------------------
 2 files changed, 7 insertions(+), 22 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c
index 92bb0ee338..20fe8dbf17 100644
--- a/src/mesa/drivers/dri/r300/r300_draw.c
+++ b/src/mesa/drivers/dri/r300/r300_draw.c
@@ -442,8 +442,6 @@ static GLboolean r300TryDrawPrims(GLcontext *ctx,
 	return GL_TRUE;
 }
 
-/* TODO: rebase if number of indices in any of primitives is > 8192 for 32bit indices or 16384 for 16bit indices */
-
 static void r300DrawPrims(GLcontext *ctx,
 			 const struct gl_client_array *arrays[],
 			 const struct _mesa_prim *prim,
@@ -455,7 +453,11 @@ static void r300DrawPrims(GLcontext *ctx,
 	struct split_limits limits;
 	GLboolean retval;
 
-	limits.max_verts = 65535;
+	if (ib)
+		limits.max_verts = 0xffffffff;
+	else
+		limits.max_verts = 65535;
+
 	limits.max_indices = 65535;
 	limits.max_vb_size = 1024*1024;
 
@@ -463,7 +465,7 @@ static void r300DrawPrims(GLcontext *ctx,
 		vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, min_index, max_index, r300DrawPrims );
 		return;
 	}
-	if ((ib && ib->count > 65536)) {
+	if ((ib && ib->count > 65535)) {
 		vbo_split_prims (ctx, arrays, prim, nr_prims, ib, min_index, max_index, r300DrawPrims, &limits);
 		return;
 	}
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index bf50b062f6..36c5ca74ab 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -359,31 +359,14 @@ void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim)
 	if (type < 0 || num_verts <= 0)
 		return;
 
-	/* Make space for at least 64 dwords.
+	/* Make space for at least 128 dwords.
 	 * This is supposed to ensure that we can get all rendering
 	 * commands into a single command buffer.
 	 */
 	rcommonEnsureCmdBufSpace(&rmesa->radeon, 128, __FUNCTION__);
 
 	if (rmesa->ind_buf.ptr) {
-		if (num_verts > 65535) {
-			/* not implemented yet */
-			WARN_ONCE("Too many elts\n");
-			return;
-		}
-		/* Note: The following is incorrect, but it's the best I can do
-		 * without a major refactoring of how DMA memory is handled.
-		 * The problem: Ensuring that both vertex arrays *and* index
-		 * arrays are at the right position, and then ensuring that
-		 * the LOAD_VBPNTR, DRAW_INDX and INDX_BUFFER packets are emitted
-		 * at once.
-		 *
-		 * So why is the following incorrect? Well, it seems like
-		 * allocating the index array might actually evict the vertex
-		 * arrays. *sigh*
-		 */
 		r300EmitElts(ctx, num_verts);
-		/* don't pass start if we are split up */
 		r300EmitAOS(rmesa, rmesa->radeon.tcl.aos_count, 0);
 		if (rmesa->radeon.radeonScreen->kernel_mm) {
 			BEGIN_BATCH_NO_AUTOSTATE(2);
-- 
cgit v1.2.3


From c27f21f92d2cf9d23a9edb15d144eceb9ff383bc Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Mon, 6 Jul 2009 14:15:00 +1000
Subject: radeon/r200/r300: port to new space checking code in libdrm

This moves a big chunk of the space checking code into libdrm so
it can be shared by the DDX.
---
 src/mesa/drivers/dri/r200/Makefile                |   8 +-
 src/mesa/drivers/dri/r200/r200_cmdbuf.c           |   8 +-
 src/mesa/drivers/dri/r200/r200_state.c            |  24 +--
 src/mesa/drivers/dri/r300/Makefile                |  10 +-
 src/mesa/drivers/dri/r300/r300_ioctl.c            |  48 ++++-
 src/mesa/drivers/dri/r300/r300_texstate.c         |  31 +--
 src/mesa/drivers/dri/radeon/Makefile              |   7 +-
 src/mesa/drivers/dri/radeon/radeon_bo_drm.h       |   8 +
 src/mesa/drivers/dri/radeon/radeon_bo_legacy.c    |  17 +-
 src/mesa/drivers/dri/radeon/radeon_common.c       |  52 +----
 src/mesa/drivers/dri/radeon/radeon_common.h       |   4 -
 src/mesa/drivers/dri/radeon/radeon_cs_drm.h       |  56 ++++--
 src/mesa/drivers/dri/radeon/radeon_cs_legacy.c    | 100 ---------
 src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c | 234 ++++++++++++++++++++++
 src/mesa/drivers/dri/radeon/radeon_dma.c          |   8 +-
 src/mesa/drivers/dri/radeon/radeon_state.c        |  25 ++-
 16 files changed, 417 insertions(+), 223 deletions(-)
 create mode 100644 src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r200/Makefile b/src/mesa/drivers/dri/r200/Makefile
index 6a246edf7c..4605f72a61 100644
--- a/src/mesa/drivers/dri/r200/Makefile
+++ b/src/mesa/drivers/dri/r200/Makefile
@@ -13,6 +13,10 @@ ifeq ($(USING_EGL), 1)
 EGL_SOURCES = server/radeon_egl.c
 endif
 
+ifeq ($(RADEON_LDFLAGS),)
+CS_SOURCES = radeon_cs_space_drm.c
+endif
+
 RADEON_COMMON_SOURCES = \
 	radeon_texture.c \
 	radeon_common_context.c \
@@ -43,6 +47,7 @@ DRIVER_SOURCES = r200_context.c \
 		 radeon_screen.c \
 		 $(EGL_SOURCES) \
 		 $(RADEON_COMMON_SOURCES)
+		 $(CS_SOURCES)
 
 C_SOURCES = $(COMMON_SOURCES) $(DRIVER_SOURCES)
 
@@ -82,7 +87,8 @@ COMMON_SYMLINKS = \
 	radeon_texture.h \
 	radeon_dma.c \
 	radeon_dma.h \
-	radeon_fbo.c
+	radeon_fbo.c \
+	$(CS_SOURCES)
 
 DRI_LIB_DEPS += $(RADEON_LDFLAGS)
 
diff --git a/src/mesa/drivers/dri/r200/r200_cmdbuf.c b/src/mesa/drivers/dri/r200/r200_cmdbuf.c
index df9dd83344..25e30eda52 100644
--- a/src/mesa/drivers/dri/r200/r200_cmdbuf.c
+++ b/src/mesa/drivers/dri/r200/r200_cmdbuf.c
@@ -225,6 +225,7 @@ GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa,
 				    GLuint min_nr )
 {
    GLushort *retval;
+   int ret;
 
    if (R200_DEBUG & DEBUG_IOCTL)
       fprintf(stderr, "%s %d prim %x\n", __FUNCTION__, min_nr, primitive);
@@ -239,10 +240,11 @@ GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa,
    rmesa->radeon.tcl.elt_dma_offset = 0;
    rmesa->tcl.elt_used = min_nr * 2;
 
-   radeon_validate_bo(&rmesa->radeon, rmesa->radeon.tcl.elt_dma_bo,
-                      RADEON_GEM_DOMAIN_GTT, 0);
-   if (radeon_revalidate_bos(rmesa->radeon.glCtx) == GL_FALSE)
+   ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, rmesa->radeon.tcl.elt_dma_bo,
+				 RADEON_GEM_DOMAIN_GTT, 0);
+   if (ret) {
       fprintf(stderr,"failure to revalidate BOs - badness\n");
+   }
 
    radeon_bo_map(rmesa->radeon.tcl.elt_dma_bo, 1);
    retval = rmesa->radeon.tcl.elt_dma_bo->ptr + rmesa->radeon.tcl.elt_dma_offset;
diff --git a/src/mesa/drivers/dri/r200/r200_state.c b/src/mesa/drivers/dri/r200/r200_state.c
index f8ebe0df57..4426f3c3ec 100644
--- a/src/mesa/drivers/dri/r200/r200_state.c
+++ b/src/mesa/drivers/dri/r200/r200_state.c
@@ -2274,23 +2274,23 @@ static GLboolean r200ValidateBuffers(GLcontext *ctx)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    struct radeon_renderbuffer *rrb;
-   int i;
+   int i, ret;
 
-   radeon_validate_reset_bos(&rmesa->radeon);
+   radeon_cs_space_reset_bos(rmesa->radeon.cmdbuf.cs);
 
    rrb = radeon_get_colorbuffer(&rmesa->radeon);
    /* color buffer */
    if (rrb && rrb->bo) {
-     radeon_validate_bo(&rmesa->radeon, rrb->bo,
-			0, RADEON_GEM_DOMAIN_VRAM);
+     radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, rrb->bo,
+				       0, RADEON_GEM_DOMAIN_VRAM);
    }
 
    /* depth buffer */
    rrb = radeon_get_depthbuffer(&rmesa->radeon);
    /* color buffer */
    if (rrb && rrb->bo) {
-     radeon_validate_bo(&rmesa->radeon, rrb->bo,
-			0, RADEON_GEM_DOMAIN_VRAM);
+     radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, rrb->bo,
+				       0, RADEON_GEM_DOMAIN_VRAM);
    }
 
    for (i = 0; i < ctx->Const.MaxTextureImageUnits; ++i) {
@@ -2301,17 +2301,17 @@ static GLboolean r200ValidateBuffers(GLcontext *ctx)
 
       t = radeon_tex_obj(ctx->Texture.Unit[i]._Current);
       if (t->image_override && t->bo)
-	radeon_validate_bo(&rmesa->radeon, t->bo,
+	radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, t->bo,
 			   RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
       else if (t->mt->bo)
-	radeon_validate_bo(&rmesa->radeon, t->mt->bo,
+	radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, t->mt->bo,
 			   RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
    }
 
-   if (rmesa->radeon.dma.current)
-      radeon_validate_bo(&rmesa->radeon, rmesa->radeon.dma.current, RADEON_GEM_DOMAIN_GTT, 0);
-
-   return radeon_revalidate_bos(ctx);
+   ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, rmesa->radeon.dma.current, RADEON_GEM_DOMAIN_GTT, 0);
+   if (ret)
+       return GL_FALSE;
+   return GL_TRUE;
 }
 
 GLboolean r200ValidateState( GLcontext *ctx )
diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile
index bdb09624be..a77209074a 100644
--- a/src/mesa/drivers/dri/r300/Makefile
+++ b/src/mesa/drivers/dri/r300/Makefile
@@ -13,6 +13,10 @@ ifeq ($(USING_EGL), 1)
 EGL_SOURCES = server/radeon_egl.c
 endif
 
+ifeq ($(RADEON_LDFLAGS),)
+CS_SOURCES = radeon_cs_space_drm.c
+endif
+
 COMMON_SOURCES = \
 	../../common/driverfuncs.c \
 	../common/mm.c \
@@ -59,7 +63,8 @@ DRIVER_SOURCES = \
 		 r300_emit.c \
 		 r300_swtcl.c \
 		 $(RADEON_COMMON_SOURCES) \
-		 $(EGL_SOURCES)
+		 $(EGL_SOURCES) \
+		 $(CS_SOURCES)
 
 C_SOURCES = $(COMMON_SOURCES) $(DRIVER_SOURCES)
 
@@ -100,7 +105,8 @@ COMMON_SYMLINKS = \
 	radeon_mipmap_tree.h \
 	radeon_texture.c \
 	radeon_texture.h \
-	radeon_fbo.c
+	radeon_fbo.c \
+	$(CS_SOURCES)
 
 DRI_LIB_DEPS += $(RADEON_LDFLAGS)
 
diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.c b/src/mesa/drivers/dri/r300/r300_ioctl.c
index 104079b4db..4ae0b4504c 100644
--- a/src/mesa/drivers/dri/r300/r300_ioctl.c
+++ b/src/mesa/drivers/dri/r300/r300_ioctl.c
@@ -540,19 +540,50 @@ static void r300EmitClearState(GLcontext * ctx)
 	}
 }
 
-static void r300KernelClear(GLcontext *ctx, GLuint flags)
+static int r300KernelClear(GLcontext *ctx, GLuint flags)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon);
 	struct radeon_framebuffer *rfb = dPriv->driverPrivate;
 	struct radeon_renderbuffer *rrb;
 	struct radeon_renderbuffer *rrbd;
-	int bits = 0;
+	int bits = 0, ret;
 
 	/* Make sure it fits there. */
+	radeon_cs_space_reset_bos(r300->radeon.cmdbuf.cs);
+	
+	if (flags & BUFFER_BIT_COLOR0) {
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_COLOR0);
+		radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
+						  rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM);
+	}
+
+	if (flags & BUFFER_BIT_FRONT_LEFT) {
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT);
+		radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
+						  rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM);
+	}
+
+	if (flags & BUFFER_BIT_BACK_LEFT) {
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_BACK_LEFT);
+		radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
+						  rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM);
+	}
+
+	rrbd = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH);
+	if (rrbd) {
+		radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
+						  rrbd->bo, 0, RADEON_GEM_DOMAIN_VRAM);
+	}
+
+	ret = radeon_cs_space_check(r300->radeon.cmdbuf.cs);
+	if (ret)
+	  return -1;
+
 	rcommonEnsureCmdBufSpace(&r300->radeon, 421 * 3, __FUNCTION__);
 	if (flags || bits)
 		r300EmitClearState(ctx);
+
 	rrbd = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH);
 	if (rrbd && (flags & BUFFER_BIT_DEPTH))
 		bits |= CLEARBUFFER_DEPTH;
@@ -582,6 +613,7 @@ static void r300KernelClear(GLcontext *ctx, GLuint flags)
 		r300ClearBuffer(r300, bits, NULL, rrbd);
 
 	COMMIT_BATCH();
+	return 0;
 }
 
 /**
@@ -593,7 +625,7 @@ static void r300Clear(GLcontext * ctx, GLbitfield mask)
 	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon);
 	const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask);
 	GLbitfield swrast_mask = 0, tri_mask = 0;
-	int i;
+	int i, ret;
 	struct gl_framebuffer *fb = ctx->DrawBuffer;
 
 	if (RADEON_DEBUG & DEBUG_IOCTL)
@@ -643,12 +675,18 @@ static void r300Clear(GLcontext * ctx, GLbitfield mask)
 	/* SW fallback clearing */
 	swrast_mask = mask & ~tri_mask;
 
+	ret = 0;
 	if (tri_mask) {
 		if (r300->radeon.radeonScreen->kernel_mm)
 			r300UserClear(ctx, tri_mask);
-		else
-			r300KernelClear(ctx, tri_mask);
+		else {
+			/* if kernel clear fails due to size restraints fallback */
+			ret = r300KernelClear(ctx, tri_mask);
+			if (ret < 0)
+				swrast_mask |= tri_mask;
+		}
 	}
+
 	if (swrast_mask) {
 		if (RADEON_DEBUG & DEBUG_FALLBACKS)
 			fprintf(stderr, "%s: swrast clear, mask: %x\n",
diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c
index c380840a00..6f489ace7b 100644
--- a/src/mesa/drivers/dri/r300/r300_texstate.c
+++ b/src/mesa/drivers/dri/r300/r300_texstate.c
@@ -281,21 +281,24 @@ GLboolean r300ValidateBuffers(GLcontext * ctx)
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct radeon_renderbuffer *rrb;
 	int i;
+	int ret;
 
-	radeon_validate_reset_bos(&rmesa->radeon);
+	radeon_cs_space_reset_bos(rmesa->radeon.cmdbuf.cs);
 
 	rrb = radeon_get_colorbuffer(&rmesa->radeon);
 	/* color buffer */
 	if (rrb && rrb->bo) {
-		radeon_validate_bo(&rmesa->radeon, rrb->bo,
-				   0, RADEON_GEM_DOMAIN_VRAM);
+		radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+						  rrb->bo, 0,
+						  RADEON_GEM_DOMAIN_VRAM);
 	}
 
 	/* depth buffer */
 	rrb = radeon_get_depthbuffer(&rmesa->radeon);
 	if (rrb && rrb->bo) {
-		radeon_validate_bo(&rmesa->radeon, rrb->bo,
-				   0, RADEON_GEM_DOMAIN_VRAM);
+		radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+						  rrb->bo, 0,
+						  RADEON_GEM_DOMAIN_VRAM);
 	}
 	
 	for (i = 0; i < ctx->Const.MaxTextureImageUnits; ++i) {
@@ -311,17 +314,19 @@ GLboolean r300ValidateBuffers(GLcontext * ctx)
 		}
 		t = radeon_tex_obj(ctx->Texture.Unit[i]._Current);
 		if (t->image_override && t->bo)
-			radeon_validate_bo(&rmesa->radeon, t->bo,
-					   RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
-
+			radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+							  t->bo,
+							  RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
 		else if (t->mt->bo)
-			radeon_validate_bo(&rmesa->radeon, t->mt->bo,
-					   RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
+			radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+							  t->mt->bo,
+							  RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
 	}
-	if (rmesa->radeon.dma.current)
-		radeon_validate_bo(&rmesa->radeon, rmesa->radeon.dma.current, RADEON_GEM_DOMAIN_GTT, 0);
 
-	return radeon_revalidate_bos(ctx);
+	ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, rmesa->radeon.dma.current, RADEON_GEM_DOMAIN_GTT, 0);
+	if (ret)
+		return GL_FALSE;
+	return GL_TRUE;
 }
 
 void r300SetTexOffset(__DRIcontext * pDRICtx, GLint texname,
diff --git a/src/mesa/drivers/dri/radeon/Makefile b/src/mesa/drivers/dri/radeon/Makefile
index ba409ba813..b59ad68f44 100644
--- a/src/mesa/drivers/dri/radeon/Makefile
+++ b/src/mesa/drivers/dri/radeon/Makefile
@@ -10,6 +10,10 @@ LIBNAME = radeon_dri.so
 
 MINIGLX_SOURCES = server/radeon_dri.c 
 
+ifeq ($(RADEON_LDFLAGS),)
+CS_SOURCES = radeon_cs_space_drm.c
+endif
+
 RADEON_COMMON_SOURCES = \
 	radeon_texture.c \
 	radeon_common_context.c \
@@ -38,7 +42,8 @@ DRIVER_SOURCES = \
 
 C_SOURCES = \
 	$(COMMON_SOURCES) \
-	$(DRIVER_SOURCES) 
+	$(DRIVER_SOURCES) \
+	$(CS_SOURCES)
 
 DRIVER_DEFINES = -DRADEON_COMMON=0
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_drm.h b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h
index 1ed13f1795..c5c28ed066 100644
--- a/src/mesa/drivers/dri/radeon/radeon_bo_drm.h
+++ b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h
@@ -68,6 +68,7 @@ struct radeon_bo_funcs {
     int (*bo_map)(struct radeon_bo *bo, int write);
     int (*bo_unmap)(struct radeon_bo *bo);
     int (*bo_wait)(struct radeon_bo *bo);
+    int (*bo_is_static)(struct radeon_bo *bo);
 };
 
 struct radeon_bo_manager {
@@ -164,6 +165,13 @@ static inline int _radeon_bo_wait(struct radeon_bo *bo,
     return bo->bom->funcs->bo_wait(bo);
 }
 
+static inline int radeon_bo_is_static(struct radeon_bo *bo)
+{
+    if (bo->bom->funcs->bo_is_static)
+	return bo->bom->funcs->bo_is_static(bo);
+    return 0;
+}
+
 #define radeon_bo_open(bom, h, s, a, d, f)\
     _radeon_bo_open(bom, h, s, a, d, f, __FILE__, __FUNCTION__, __LINE__)
 #define radeon_bo_ref(bo)\
diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c
index 6a8da402b1..1fec7444cd 100644
--- a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c
+++ b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c
@@ -528,12 +528,21 @@ static int bo_unmap(struct radeon_bo *bo)
     return 0;
 }
 
+
+static int bo_is_static(struct radeon_bo *bo)
+{
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+    return bo_legacy->static_bo;
+}
+
 static struct radeon_bo_funcs bo_legacy_funcs = {
     bo_open,
     bo_ref,
     bo_unref,
     bo_map,
-    bo_unmap
+    bo_unmap,
+    NULL,
+    bo_is_static,
 };
 
 static int bo_vram_validate(struct radeon_bo *bo,
@@ -822,9 +831,3 @@ unsigned radeon_bo_legacy_relocs_size(struct radeon_bo *bo)
     return bo->size;
 }
 
-int radeon_legacy_bo_is_static(struct radeon_bo *bo)
-{
-    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
-    return bo_legacy->static_bo;
-}
-
diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c
index 7bd4a6f14f..bac72fbdfa 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common.c
+++ b/src/mesa/drivers/dri/radeon/radeon_common.c
@@ -964,58 +964,17 @@ static INLINE void radeonEmitAtoms(radeonContextPtr radeon, GLboolean dirty)
 	COMMIT_BATCH();
 }
 
-GLboolean radeon_revalidate_bos(GLcontext *ctx)
+static GLboolean radeon_revalidate_bos(GLcontext *ctx)
 {
 	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
-	int flushed = 0;
 	int ret;
-again:
-	ret = radeon_cs_space_check(radeon->cmdbuf.cs, radeon->state.bos, radeon->state.validated_bo_count);
-	if (ret == RADEON_CS_SPACE_OP_TO_BIG)
+
+	ret = radeon_cs_space_check(radeon->cmdbuf.cs);
+	if (ret == RADEON_CS_SPACE_FLUSH)
 		return GL_FALSE;
-	if (ret == RADEON_CS_SPACE_FLUSH) {
-		radeonFlush(ctx);
-		if (flushed)
-			return GL_FALSE;
-		flushed = 1;
-		goto again;
-	}
 	return GL_TRUE;
 }
 
-void radeon_validate_reset_bos(radeonContextPtr radeon)
-{
-	int i;
-
-	for (i = 0; i < radeon->state.validated_bo_count; i++) {
-		radeon_bo_unref(radeon->state.bos[i].bo);
-		radeon->state.bos[i].bo = NULL;
-		radeon->state.bos[i].read_domains = 0;
-		radeon->state.bos[i].write_domain = 0;
-		radeon->state.bos[i].new_accounted = 0;
-	}
-	radeon->state.validated_bo_count = 0;
-}
-
-void radeon_validate_bo(radeonContextPtr radeon, struct radeon_bo *bo, uint32_t read_domains, uint32_t write_domain)
-{
-	int i;
-	for (i = 0; i < radeon->state.validated_bo_count; i++) {
-		if (radeon->state.bos[i].bo == bo &&
-		    radeon->state.bos[i].read_domains == read_domains &&
-		    radeon->state.bos[i].write_domain == write_domain)
-			return;
-	}
-	radeon_bo_ref(bo);
-	radeon->state.bos[radeon->state.validated_bo_count].bo = bo;
-	radeon->state.bos[radeon->state.validated_bo_count].read_domains = read_domains;
-	radeon->state.bos[radeon->state.validated_bo_count].write_domain = write_domain;
-	radeon->state.bos[radeon->state.validated_bo_count].new_accounted = 0;
-	radeon->state.validated_bo_count++;
-
-	assert(radeon->state.validated_bo_count < RADEON_MAX_BOS);
-}
-
 void radeonEmitState(radeonContextPtr radeon)
 {
 	if (RADEON_DEBUG & (DEBUG_STATE|DEBUG_PRIMS))
@@ -1229,6 +1188,9 @@ void rcommonInitCmdBuf(radeonContextPtr rmesa)
 	assert(rmesa->cmdbuf.cs != NULL);
 	rmesa->cmdbuf.size = size;
 
+	radeon_cs_space_set_flush(rmesa->cmdbuf.cs,
+				  (void (*)(void *))radeonFlush, rmesa->glCtx);
+
 	if (!rmesa->radeonScreen->kernel_mm) {
 		radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, rmesa->radeonScreen->texSize[0]);
 		radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, rmesa->radeonScreen->gartTextures.size);
diff --git a/src/mesa/drivers/dri/radeon/radeon_common.h b/src/mesa/drivers/dri/radeon/radeon_common.h
index b60792df0b..ba6c7c5773 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common.h
+++ b/src/mesa/drivers/dri/radeon/radeon_common.h
@@ -47,10 +47,6 @@ void radeon_get_cliprects(radeonContextPtr radeon,
 			  struct drm_clip_rect **cliprects,
 			  unsigned int *num_cliprects,
 			  int *x_off, int *y_off);
-GLboolean radeon_revalidate_bos(GLcontext *ctx);
-void radeon_validate_bo(radeonContextPtr radeon, struct radeon_bo *bo, uint32_t read_domains, uint32_t write_domain);
-void radeon_validate_reset_bos(radeonContextPtr radeon);
-
 void radeon_fbo_init(struct radeon_context *radeon);
 void
 radeon_renderbuffer_set_bo(struct radeon_renderbuffer *rb,
diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_drm.h b/src/mesa/drivers/dri/radeon/radeon_cs_drm.h
index 984725a6c9..ee403d173c 100644
--- a/src/mesa/drivers/dri/radeon/radeon_cs_drm.h
+++ b/src/mesa/drivers/dri/radeon/radeon_cs_drm.h
@@ -56,6 +56,8 @@ struct radeon_cs_space_check {
     uint32_t new_accounted;
 };
 
+#define MAX_SPACE_BOS (32)
+
 struct radeon_cs_manager;
 
 struct radeon_cs {
@@ -72,7 +74,10 @@ struct radeon_cs {
     const char                  *section_file;
     const char                  *section_func;
     int                         section_line;
-
+    struct radeon_cs_space_check bos[MAX_SPACE_BOS];
+    int                         bo_count;
+    void                        (*space_flush_fn)(void *);
+    void                        *space_flush_data;
 };
 
 /* cs functions */
@@ -98,16 +103,14 @@ struct radeon_cs_funcs {
     int (*cs_erase)(struct radeon_cs *cs);
     int (*cs_need_flush)(struct radeon_cs *cs);
     void (*cs_print)(struct radeon_cs *cs, FILE *file);
-    int (*cs_space_check)(struct radeon_cs *cs, struct radeon_cs_space_check *bos,
-			  int num_bo);
 };
 
 struct radeon_cs_manager {
     struct radeon_cs_funcs  *funcs;
     int                     fd;
-    uint32_t vram_limit, gart_limit;
-    uint32_t vram_write_used, gart_write_used;
-    uint32_t read_used;
+    int32_t vram_limit, gart_limit;
+    int32_t vram_write_used, gart_write_used;
+    int32_t read_used;
 };
 
 static inline struct radeon_cs *radeon_cs_create(struct radeon_cs_manager *csm,
@@ -171,13 +174,6 @@ static inline void radeon_cs_print(struct radeon_cs *cs, FILE *file)
     cs->csm->funcs->cs_print(cs, file);
 }
 
-static inline int radeon_cs_space_check(struct radeon_cs *cs,
-					    struct radeon_cs_space_check *bos,
-					    int num_bo)
-{
-    return cs->csm->funcs->cs_space_check(cs, bos, num_bo);
-}
-
 static inline void radeon_cs_set_limit(struct radeon_cs *cs, uint32_t domain, uint32_t limit)
 {
     
@@ -204,4 +200,38 @@ static inline void radeon_cs_write_qword(struct radeon_cs *cs, uint64_t qword)
         cs->section_cdw+=2;
     }
 }
+
+static inline void radeon_cs_space_set_flush(struct radeon_cs *cs, void (*fn)(void *), void *data)
+{
+    cs->space_flush_fn = fn;
+    cs->space_flush_data = data;
+}
+
+
+/*
+ * add a persistent BO to the list
+ * a persistent BO is one that will be referenced across flushes,
+ * i.e. colorbuffer, textures etc.
+ * They get reset when a new "operation" happens, where an operation
+ * is a state emission with a color/textures etc followed by a bunch of vertices.
+ */
+void radeon_cs_space_add_persistent_bo(struct radeon_cs *cs,
+				       struct radeon_bo *bo,
+				       uint32_t read_domains,
+				       uint32_t write_domain);
+
+/* reset the persistent BO list */
+void radeon_cs_space_reset_bos(struct radeon_cs *cs);
+
+/* do a space check with the current persistent BO list */
+int radeon_cs_space_check(struct radeon_cs *cs);
+
+/* do a space check with the current persistent BO list and a temporary BO
+ * a temporary BO is like a DMA buffer, which  gets flushed with the
+ * command buffer */
+int radeon_cs_space_check_with_bo(struct radeon_cs *cs,
+				  struct radeon_bo *bo,
+				  uint32_t read_domains,
+				  uint32_t write_domain);
+
 #endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c
index ac94789417..f77e22c437 100644
--- a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c
+++ b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c
@@ -365,105 +365,6 @@ static void cs_print(struct radeon_cs *cs, FILE *file)
 {
 }
 
-static int cs_check_space(struct radeon_cs *cs, struct radeon_cs_space_check *bos, int num_bo)
-{
-    struct radeon_cs_manager *csm = cs->csm;
-    int this_op_read = 0, this_op_gart_write = 0, this_op_vram_write = 0;
-    uint32_t read_domains, write_domain;
-    int i;
-    struct radeon_bo *bo;
-
-    /* check the totals for this operation */
-
-    if (num_bo == 0)
-        return 0;
-
-    /* prepare */
-    for (i = 0; i < num_bo; i++) {
-      bo = bos[i].bo;
-
-      bos[i].new_accounted = 0;
-      read_domains = bos[i].read_domains;
-      write_domain = bos[i].write_domain;
-		
-      /* pinned bos don't count */
-      if (radeon_legacy_bo_is_static(bo))
-	  continue;
- 
-      /* already accounted this bo */
-      if (write_domain && (write_domain == bo->space_accounted)) {
-	  bos[i].new_accounted = bo->space_accounted;
-	  continue;
-      }
-
-      if (read_domains && ((read_domains << 16) == bo->space_accounted)) {
-	  bos[i].new_accounted = bo->space_accounted;
-	  continue;
-      }
-      
-      if (bo->space_accounted == 0) {
-	  if (write_domain == RADEON_GEM_DOMAIN_VRAM)
-	      this_op_vram_write += bo->size;
-	  else if (write_domain == RADEON_GEM_DOMAIN_GTT)
-	      this_op_gart_write += bo->size;
-	  else
-	      this_op_read += bo->size;
-	  bos[i].new_accounted = (read_domains << 16) | write_domain;
-      } else {
-	  uint16_t old_read, old_write;
-	  
-	  old_read = bo->space_accounted >> 16;
-	  old_write = bo->space_accounted & 0xffff;
-
-	  if (write_domain && (old_read & write_domain)) {
-	      bos[i].new_accounted = write_domain;
-	      /* moving from read to a write domain */
-	      if (write_domain == RADEON_GEM_DOMAIN_VRAM) {
-		  this_op_read -= bo->size;
-		  this_op_vram_write += bo->size;
-	      } else if (write_domain == RADEON_GEM_DOMAIN_VRAM) {
-		  this_op_read -= bo->size;
-		  this_op_gart_write += bo->size;
-	      }
-	  } else if (read_domains & old_write) {
-	      bos[i].new_accounted = bo->space_accounted & 0xffff;
-	  } else {
-	      /* rewrite the domains */
-	      if (write_domain != old_write)
-		  fprintf(stderr,"WRITE DOMAIN RELOC FAILURE 0x%x %d %d\n", bo->handle, write_domain, old_write);
-	      if (read_domains != old_read)
-		  fprintf(stderr,"READ DOMAIN RELOC FAILURE 0x%x %d %d\n", bo->handle, read_domains, old_read);
-	      return RADEON_CS_SPACE_FLUSH;
-	  }
-      }
-	}
-	
-	if (this_op_read < 0)
-		this_op_read = 0;
-
-	/* check sizes - operation first */
-	if ((this_op_read + this_op_gart_write > csm->gart_limit) ||
-	    (this_op_vram_write > csm->vram_limit)) {
-	    return RADEON_CS_SPACE_OP_TO_BIG;
-	}
-
-	if (((csm->vram_write_used + this_op_vram_write) > csm->vram_limit) ||
-	    ((csm->read_used + csm->gart_write_used + this_op_gart_write + this_op_read) > csm->gart_limit)) {
-		return RADEON_CS_SPACE_FLUSH;
-	}
-
-	csm->gart_write_used += this_op_gart_write;
-	csm->vram_write_used += this_op_vram_write;
-	csm->read_used += this_op_read;
-	/* commit */
-	for (i = 0; i < num_bo; i++) {
-		bo = bos[i].bo;
-		bo->space_accounted = bos[i].new_accounted;
-	}
-
-	return RADEON_CS_SPACE_OK;
-}
-
 static struct radeon_cs_funcs  radeon_cs_legacy_funcs = {
     cs_create,
     cs_write_reloc,
@@ -474,7 +375,6 @@ static struct radeon_cs_funcs  radeon_cs_legacy_funcs = {
     cs_erase,
     cs_need_flush,
     cs_print,
-    cs_check_space
 };
 
 struct radeon_cs_manager *radeon_cs_manager_legacy_ctor(struct radeon_context *ctx)
diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c b/src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c
new file mode 100644
index 0000000000..5a8df7bb8c
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c
@@ -0,0 +1,234 @@
+/* 
+ * Copyright © 2009 Red Hat Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
+ * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ */
+#include <assert.h>
+#include <errno.h>
+#include <stdlib.h>
+#include "radeon_bocs_wrapper.h"
+
+struct rad_sizes {
+    int32_t op_read;
+    int32_t op_gart_write;
+    int32_t op_vram_write;
+};
+
+static inline int radeon_cs_setup_bo(struct radeon_cs_space_check *sc, struct rad_sizes *sizes)
+{
+    uint32_t read_domains, write_domain;
+    struct radeon_bo *bo;
+
+    bo = sc->bo;
+    sc->new_accounted = 0;
+    read_domains = sc->read_domains;
+    write_domain = sc->write_domain;
+
+    /* legacy needs a static check */
+    if (radeon_bo_is_static(bo)) {
+	bo->space_accounted = sc->new_accounted = (read_domains << 16) | write_domain;
+	return 0;
+    }
+
+    /* already accounted this bo */
+    if (write_domain && (write_domain == bo->space_accounted)) {
+	sc->new_accounted = bo->space_accounted;
+	return 0;
+    }
+    if (read_domains && ((read_domains << 16) == bo->space_accounted)) {
+	sc->new_accounted = bo->space_accounted;
+	return 0;
+    }
+
+    if (bo->space_accounted == 0) {
+	if (write_domain == RADEON_GEM_DOMAIN_VRAM)
+	    sizes->op_vram_write += bo->size;
+	else if (write_domain == RADEON_GEM_DOMAIN_GTT)
+	  sizes->op_gart_write += bo->size;
+	else
+	    sizes->op_read += bo->size;
+	sc->new_accounted = (read_domains << 16) | write_domain;
+    } else {
+	uint16_t old_read, old_write;
+	
+	old_read = bo->space_accounted >> 16;
+	old_write = bo->space_accounted & 0xffff;
+	
+	if (write_domain && (old_read & write_domain)) {
+	    sc->new_accounted = write_domain;
+	    /* moving from read to a write domain */
+	    if (write_domain == RADEON_GEM_DOMAIN_VRAM) {
+		sizes->op_read -= bo->size;
+		sizes->op_vram_write += bo->size;
+	    } else if (write_domain == RADEON_GEM_DOMAIN_VRAM) {
+		sizes->op_read -= bo->size;
+		sizes->op_gart_write += bo->size;
+	    }
+	} else if (read_domains & old_write) {
+	    sc->new_accounted = bo->space_accounted & 0xffff;
+	} else {
+	    /* rewrite the domains */
+	    if (write_domain != old_write)
+		fprintf(stderr,"WRITE DOMAIN RELOC FAILURE 0x%x %d %d\n", bo->handle, write_domain, old_write);
+	    if (read_domains != old_read)
+		fprintf(stderr,"READ DOMAIN RELOC FAILURE 0x%x %d %d\n", bo->handle, read_domains, old_read);
+	    return RADEON_CS_SPACE_FLUSH;
+	}
+    }
+    return 0;
+}
+
+static int radeon_cs_do_space_check(struct radeon_cs *cs, struct radeon_cs_space_check *new_tmp)
+{
+    struct radeon_cs_manager *csm = cs->csm;
+    int i;
+    struct radeon_bo *bo;
+    struct rad_sizes sizes;
+    int ret;
+
+    /* check the totals for this operation */
+
+    if (cs->bo_count == 0 && !new_tmp)
+	return 0;
+
+    memset(&sizes, 0, sizeof(struct rad_sizes));
+
+    /* prepare */
+    for (i = 0; i < cs->bo_count; i++) {
+	ret = radeon_cs_setup_bo(&cs->bos[i], &sizes);
+	if (ret)
+	    return ret;
+    }
+
+    if (new_tmp) {
+	ret = radeon_cs_setup_bo(new_tmp, &sizes);
+	if (ret)
+	    return ret;
+    }
+	
+    if (sizes.op_read < 0)
+	    sizes.op_read = 0;
+
+    /* check sizes - operation first */
+    if ((sizes.op_read + sizes.op_gart_write > csm->gart_limit) ||
+	(sizes.op_vram_write > csm->vram_limit)) {
+	    return RADEON_CS_SPACE_OP_TO_BIG;
+    }
+    
+    if (((csm->vram_write_used + sizes.op_vram_write) > csm->vram_limit) ||
+	((csm->read_used + csm->gart_write_used + sizes.op_gart_write + sizes.op_read) > csm->gart_limit)) {
+	    return RADEON_CS_SPACE_FLUSH;
+    }
+    
+    csm->gart_write_used += sizes.op_gart_write;
+    csm->vram_write_used += sizes.op_vram_write;
+    csm->read_used += sizes.op_read;
+    /* commit */
+    for (i = 0; i < cs->bo_count; i++) {
+	    bo = cs->bos[i].bo;
+	    bo->space_accounted = cs->bos[i].new_accounted;
+    }
+    if (new_tmp)
+	new_tmp->bo->space_accounted = new_tmp->new_accounted;
+    
+    return RADEON_CS_SPACE_OK;
+}
+
+void radeon_cs_space_add_persistent_bo(struct radeon_cs *cs, struct radeon_bo *bo, uint32_t read_domains, uint32_t write_domain)
+{
+    int i;
+    for (i = 0; i < cs->bo_count; i++) {
+	if (cs->bos[i].bo == bo &&
+	    cs->bos[i].read_domains == read_domains &&
+	    cs->bos[i].write_domain == write_domain)
+	    return;
+    }
+    radeon_bo_ref(bo);
+    i = cs->bo_count;
+    cs->bos[i].bo = bo;
+    cs->bos[i].read_domains = read_domains;
+    cs->bos[i].write_domain = write_domain;
+    cs->bos[i].new_accounted = 0;
+    cs->bo_count++;
+
+    assert(cs->bo_count < MAX_SPACE_BOS);
+}
+
+static int radeon_cs_check_space_internal(struct radeon_cs *cs, struct radeon_cs_space_check *tmp_bo)
+{
+    int ret;
+    int flushed = 0;
+
+again:
+    ret = radeon_cs_do_space_check(cs, tmp_bo);
+    if (ret == RADEON_CS_SPACE_OP_TO_BIG)
+	return -1;
+    if (ret == RADEON_CS_SPACE_FLUSH) {
+	(*cs->space_flush_fn)(cs->space_flush_data);
+	if (flushed)
+	    return -1;
+	flushed = 1;
+	goto again;
+    }
+    return 0;
+}
+
+int radeon_cs_space_check_with_bo(struct radeon_cs *cs,
+				  struct radeon_bo *bo,
+				  uint32_t read_domains, uint32_t write_domain)
+{									
+    struct radeon_cs_space_check temp_bo;
+    int ret = 0;
+
+    if (bo) {
+	temp_bo.bo = bo;
+	temp_bo.read_domains = read_domains;
+	temp_bo.write_domain = write_domain;
+	temp_bo.new_accounted = 0;
+    }
+
+    ret = radeon_cs_check_space_internal(cs, bo ? &temp_bo : NULL);
+    return ret;
+}
+
+int radeon_cs_space_check(struct radeon_cs *cs)
+{
+    return radeon_cs_check_space_internal(cs, NULL);
+}
+
+void radeon_cs_space_reset_bos(struct radeon_cs *cs)
+{
+    int i;
+    for (i = 0; i < cs->bo_count; i++) {
+	radeon_bo_unref(cs->bos[i].bo);
+	cs->bos[i].bo = NULL;
+	cs->bos[i].read_domains = 0;
+	cs->bos[i].write_domain = 0;
+	cs->bos[i].new_accounted = 0;
+    }
+    cs->bo_count = 0;
+}
+
+
diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c
index 48b0d63818..dc9a0157f7 100644
--- a/src/mesa/drivers/dri/radeon/radeon_dma.c
+++ b/src/mesa/drivers/dri/radeon/radeon_dma.c
@@ -198,10 +198,10 @@ again_alloc:
 	rmesa->dma.current_used = 0;
 	rmesa->dma.current_vertexptr = 0;
 	
-	radeon_validate_bo(rmesa, rmesa->dma.current, RADEON_GEM_DOMAIN_GTT, 0);
-
-	if (radeon_revalidate_bos(rmesa->glCtx) == GL_FALSE)
-	  fprintf(stderr,"failure to revalidate BOs - badness\n");
+	if (radeon_cs_space_check_with_bo(rmesa->cmdbuf.cs,
+					  rmesa->dma.current,
+					  RADEON_GEM_DOMAIN_GTT, 0))
+		fprintf(stderr,"failure to revalidate BOs - badness\n");
 
 	if (!rmesa->dma.current) {
         /* Cmd buff have been flushed in radeon_revalidate_bos */
diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c
index 06b8c29936..528348a300 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state.c
@@ -2048,23 +2048,23 @@ static GLboolean r100ValidateBuffers(GLcontext *ctx)
 {
    r100ContextPtr rmesa = R100_CONTEXT(ctx);
    struct radeon_renderbuffer *rrb;
-   int i;
+   int i, ret;
 
-   radeon_validate_reset_bos(&rmesa->radeon);
+   radeon_cs_space_reset_bos(rmesa->radeon.cmdbuf.cs);
 
    rrb = radeon_get_colorbuffer(&rmesa->radeon);
    /* color buffer */
    if (rrb && rrb->bo) {
-     radeon_validate_bo(&rmesa->radeon, rrb->bo,
-			0, RADEON_GEM_DOMAIN_VRAM);
+     radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, rrb->bo,
+				       0, RADEON_GEM_DOMAIN_VRAM);
    }
 
    /* depth buffer */
    rrb = radeon_get_depthbuffer(&rmesa->radeon);
    /* color buffer */
    if (rrb && rrb->bo) {
-     radeon_validate_bo(&rmesa->radeon, rrb->bo,
-			0, RADEON_GEM_DOMAIN_VRAM);
+     radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, rrb->bo,
+				       0, RADEON_GEM_DOMAIN_VRAM);
    }
 
    for (i = 0; i < ctx->Const.MaxTextureImageUnits; ++i) {
@@ -2075,18 +2075,17 @@ static GLboolean r100ValidateBuffers(GLcontext *ctx)
 
       t = radeon_tex_obj(ctx->Texture.Unit[i]._Current);
       if (t->image_override && t->bo)
-	radeon_validate_bo(&rmesa->radeon, t->bo,
+	radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, t->bo,
 			   RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
       else if (t->mt->bo)
-	radeon_validate_bo(&rmesa->radeon, t->mt->bo,
+	radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, t->mt->bo,
 			   RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
    }
 
-   if (rmesa->radeon.dma.current)
-       radeon_validate_bo(&rmesa->radeon, rmesa->radeon.dma.current,
-			  RADEON_GEM_DOMAIN_GTT, 0);
-
-   return radeon_revalidate_bos(ctx);
+   ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, rmesa->radeon.dma.current, RADEON_GEM_DOMAIN_GTT, 0);
+   if (ret)
+       return GL_FALSE;
+   return GL_TRUE;
 }
 
 GLboolean radeonValidateState( GLcontext *ctx )
-- 
cgit v1.2.3


From d66de6c341a859a30ef010c527f9a0c5865b5d65 Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Mon, 6 Jul 2009 20:35:15 +1000
Subject: radeon: ensure cmdbuf space for state + AOS is available

The problem is if we find out later we don't have any cmdbuf space but
we've already written the arrays to the DMA buffer object, we end up
emitting the current cmdbuf which has references to the current DMA object
we then send that to the hw and we can't reference the arrays we just emitted
to the old DMA buffer. things go bad, crash boom.

This can probably be tuned further + swtcl probably needs some fixes
---
 src/mesa/drivers/dri/r300/r300_draw.c | 6 ++++++
 1 file changed, 6 insertions(+)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c
index 20fe8dbf17..fe8014b358 100644
--- a/src/mesa/drivers/dri/r300/r300_draw.c
+++ b/src/mesa/drivers/dri/r300/r300_draw.c
@@ -417,6 +417,12 @@ static GLboolean r300TryDrawPrims(GLcontext *ctx,
 
 	r300FixupIndexBuffer(ctx, ib, bo, &nr_bo);
 
+	/* ensure we have the cmd buf space in advance to cover
+ 	 * the state + DMA AOS pointers */
+	rcommonEnsureCmdBufSpace(&r300->radeon,
+                           r300->radeon.hw.max_state_size + (50*sizeof(int)),
+                           __FUNCTION__);
+
 	r300SetVertexFormat(ctx, arrays, max_index + 1, bo, &nr_bo);
 
 	if (r300->fallback)
-- 
cgit v1.2.3


From 186d187ff7eebad58a18721878c5b23799e3f3d6 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 4 Jul 2009 11:33:47 +0200
Subject: r300: fix regression introduced by
 ca13937ef97c7779f639dcfc95b3798a11de01bd

Stride == 0 means that we value for first vertex should be copied to every other vertices (e.g. constant color).

This fixes glean/vertProg1 and sauerbraten with enabled shaders.
---
 src/mesa/drivers/dri/r300/r300_draw.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c
index fe8014b358..4e8b62f186 100644
--- a/src/mesa/drivers/dri/r300/r300_draw.c
+++ b/src/mesa/drivers/dri/r300/r300_draw.c
@@ -234,7 +234,7 @@ static void r300TranslateAttrib(GLcontext *ctx, GLuint attr, int count, const st
 		type = input->Type;
 		r300_attr.free_needed = GL_FALSE;
 		r300_attr.data = (GLvoid *)src_ptr;
-		r300_attr.stride = stride;
+		r300_attr.stride = input->StrideB;
 		r300_attr.dwords = (getTypeSize(type) * input->Size  + 3)/ 4;
 	}
 
-- 
cgit v1.2.3


From dfecf217fa0d7677bdd3445e32a7bb0d022a599b Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Sun, 12 Jul 2009 21:35:24 +1000
Subject: r300: fix clear mask to not use sw if not necessary

---
 src/mesa/drivers/dri/r300/r300_ioctl.c | 2 ++
 1 file changed, 2 insertions(+)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.c b/src/mesa/drivers/dri/r300/r300_ioctl.c
index 4ae0b4504c..4e913dba29 100644
--- a/src/mesa/drivers/dri/r300/r300_ioctl.c
+++ b/src/mesa/drivers/dri/r300/r300_ioctl.c
@@ -646,6 +646,8 @@ static void r300Clear(GLcontext * ctx, GLbitfield mask)
 
 	if (colorMask == ~0)
 	  tri_mask |= (mask & BUFFER_BITS_COLOR);
+	else
+	  tri_mask |= (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT));
 
 
 	/* HW stencil */
-- 
cgit v1.2.3


From f030e2ba17a3b859d30017cfd990552d3af4bad3 Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Sun, 12 Jul 2009 21:35:59 +1000
Subject: r300: move fallback warnings inside fallback debugging

random output is bad
---
 src/mesa/drivers/dri/r300/r300_render.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index 36c5ca74ab..08d67b73ed 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -452,7 +452,8 @@ void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode)
 	
 	if (mode) {
 		if ((fallback_warn & bit) == 0) {
-			_mesa_fprintf(stderr, "WARNING! Falling back to software for %s\n", getFallbackString(bit));
+			if (RADEON_DEBUG & DEBUG_FALLBACKS)
+				_mesa_fprintf(stderr, "WARNING! Falling back to software for %s\n", getFallbackString(bit));
 			fallback_warn |= bit;
 		}
 		rmesa->fallback |= bit;
-- 
cgit v1.2.3


From 9b781ca2ce9cdf6c21eb5dda4709366b2581d17a Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Mon, 29 Jun 2009 21:39:59 +0200
Subject: r300: simplify insert_wpos a little

---
 src/mesa/drivers/dri/r300/r300_vertprog.c | 76 +++++++++++++------------------
 1 file changed, 31 insertions(+), 45 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index c41a8fdd62..1d198cde42 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -1283,52 +1283,38 @@ static void insert_wpos(struct r300_vertex_program *vp, struct gl_program *prog,
 			GLuint temp_index)
 {
 	struct prog_instruction *vpi;
-	struct prog_instruction *vpi_insert;
-	int i = 0;
-
-	vpi = _mesa_alloc_instructions(prog->NumInstructions + 2);
-	_mesa_init_instructions(vpi, prog->NumInstructions + 2);
-	/* all but END */
-	_mesa_copy_instructions(vpi, prog->Instructions,
-				prog->NumInstructions - 1);
-	/* END */
-	_mesa_copy_instructions(&vpi[prog->NumInstructions + 1],
-				&prog->Instructions[prog->NumInstructions - 1],
-				1);
-	vpi_insert = &vpi[prog->NumInstructions - 1];
-
-	vpi_insert[i].Opcode = OPCODE_MOV;
-
-	vpi_insert[i].DstReg.File = PROGRAM_OUTPUT;
-	vpi_insert[i].DstReg.Index = VERT_RESULT_HPOS;
-	vpi_insert[i].DstReg.WriteMask = WRITEMASK_XYZW;
-	vpi_insert[i].DstReg.CondMask = COND_TR;
-
-	vpi_insert[i].SrcReg[0].File = PROGRAM_TEMPORARY;
-	vpi_insert[i].SrcReg[0].Index = temp_index;
-	vpi_insert[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-	i++;
-
-	vpi_insert[i].Opcode = OPCODE_MOV;
-
-	vpi_insert[i].DstReg.File = PROGRAM_OUTPUT;
-	vpi_insert[i].DstReg.Index = VERT_RESULT_TEX0 + vp->wpos_idx;
-	vpi_insert[i].DstReg.WriteMask = WRITEMASK_XYZW;
-	vpi_insert[i].DstReg.CondMask = COND_TR;
-
-	vpi_insert[i].SrcReg[0].File = PROGRAM_TEMPORARY;
-	vpi_insert[i].SrcReg[0].Index = temp_index;
-	vpi_insert[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-	i++;
 
-	free(prog->Instructions);
+	_mesa_insert_instructions(prog, prog->NumInstructions - 1, 2);
 
-	prog->Instructions = vpi;
+	vpi = &prog->Instructions[prog->NumInstructions - 3];
 
-	prog->NumInstructions += i;
-	vpi = &prog->Instructions[prog->NumInstructions - 1];
+	vpi->Opcode = OPCODE_MOV;
 
-	assert(vpi->Opcode == OPCODE_END);
+	vpi->DstReg.File = PROGRAM_OUTPUT;
+	vpi->DstReg.Index = VERT_RESULT_HPOS;
+	vpi->DstReg.WriteMask = WRITEMASK_XYZW;
+	vpi->DstReg.CondMask = COND_TR;
+
+	vpi->SrcReg[0].File = PROGRAM_TEMPORARY;
+	vpi->SrcReg[0].Index = temp_index;
+	vpi->SrcReg[0].Swizzle = SWIZZLE_XYZW;
+
+	++vpi;
+
+	vpi->Opcode = OPCODE_MOV;
+
+	vpi->DstReg.File = PROGRAM_OUTPUT;
+	vpi->DstReg.Index = VERT_RESULT_TEX0 + vp->wpos_idx;
+	vpi->DstReg.WriteMask = WRITEMASK_XYZW;
+	vpi->DstReg.CondMask = COND_TR;
+
+	vpi->SrcReg[0].File = PROGRAM_TEMPORARY;
+	vpi->SrcReg[0].Index = temp_index;
+	vpi->SrcReg[0].Swizzle = SWIZZLE_XYZW;
+
+	++vpi;
+
+	vpi->Opcode = OPCODE_END;
 }
 
 static void pos_as_texcoord(struct r300_vertex_program *vp,
@@ -1336,16 +1322,16 @@ static void pos_as_texcoord(struct r300_vertex_program *vp,
 {
 	struct prog_instruction *vpi;
 	GLuint tempregi = prog->NumTemporaries;
-	/* should do something else if no temps left... */
+
 	prog->NumTemporaries++;
 
 	for (vpi = prog->Instructions; vpi->Opcode != OPCODE_END; vpi++) {
-		if (vpi->DstReg.File == PROGRAM_OUTPUT
-		    && vpi->DstReg.Index == VERT_RESULT_HPOS) {
+		if (vpi->DstReg.File == PROGRAM_OUTPUT && vpi->DstReg.Index == VERT_RESULT_HPOS) {
 			vpi->DstReg.File = PROGRAM_TEMPORARY;
 			vpi->DstReg.Index = tempregi;
 		}
 	}
+
 	insert_wpos(vp, prog, tempregi);
 }
 
-- 
cgit v1.2.3


From 4a6899e080bbdb556926031345ef6a9fbe6d9ff2 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Mon, 29 Jun 2009 21:48:35 +0200
Subject: r300: use mesa provided function for adding MVP code

---
 src/mesa/drivers/dri/r300/r300_context.c  |   2 -
 src/mesa/drivers/dri/r300/r300_vertprog.c | 100 +++---------------------------
 2 files changed, 7 insertions(+), 95 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index b7d75426c5..6f3aab986d 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -418,8 +418,6 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 
 	_mesa_set_mvp_with_dp4( ctx, GL_TRUE );
 
-	_mesa_set_mvp_with_dp4( ctx, GL_TRUE );
-
 	/* Initialize the software rasterizer and helper modules.
 	 */
 	_swrast_CreateContext(ctx);
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index 1d198cde42..5da9bda281 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -32,6 +32,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/macros.h"
 #include "main/enums.h"
 #include "shader/program.h"
+#include "shader/programopt.h"
 #include "shader/prog_instruction.h"
 #include "shader/prog_parameter.h"
 #include "shader/prog_print.h"
@@ -1191,94 +1192,6 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp,
 	}
 }
 
-/* DP4 version seems to trigger some hw peculiarity */
-//#define PREFER_DP4
-
-static void position_invariant(struct gl_program *prog)
-{
-	struct prog_instruction *vpi;
-	struct gl_program_parameter_list *paramList;
-	int i;
-
-	gl_state_index tokens[STATE_LENGTH] = { STATE_MVP_MATRIX, 0, 0, 0, 0 };
-
-	/* tokens[4] = matrix modifier */
-#ifdef PREFER_DP4
-	tokens[4] = 0;		/* not transposed or inverted */
-#else
-	tokens[4] = STATE_MATRIX_TRANSPOSE;
-#endif
-	paramList = prog->Parameters;
-
-	vpi = _mesa_alloc_instructions(prog->NumInstructions + 4);
-	_mesa_init_instructions(vpi, prog->NumInstructions + 4);
-
-	for (i = 0; i < 4; i++) {
-		GLint idx;
-		tokens[2] = tokens[3] = i;	/* matrix row[i]..row[i] */
-		idx = _mesa_add_state_reference(paramList, tokens);
-#ifdef PREFER_DP4
-		vpi[i].Opcode = OPCODE_DP4;
-		vpi[i].StringPos = 0;
-		vpi[i].Data = 0;
-
-		vpi[i].DstReg.File = PROGRAM_OUTPUT;
-		vpi[i].DstReg.Index = VERT_RESULT_HPOS;
-		vpi[i].DstReg.WriteMask = 1 << i;
-		vpi[i].DstReg.CondMask = COND_TR;
-
-		vpi[i].SrcReg[0].File = PROGRAM_STATE_VAR;
-		vpi[i].SrcReg[0].Index = idx;
-		vpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-
-		vpi[i].SrcReg[1].File = PROGRAM_INPUT;
-		vpi[i].SrcReg[1].Index = VERT_ATTRIB_POS;
-		vpi[i].SrcReg[1].Swizzle = SWIZZLE_XYZW;
-#else
-		if (i == 0)
-			vpi[i].Opcode = OPCODE_MUL;
-		else
-			vpi[i].Opcode = OPCODE_MAD;
-
-		vpi[i].Data = 0;
-
-		if (i == 3)
-			vpi[i].DstReg.File = PROGRAM_OUTPUT;
-		else
-			vpi[i].DstReg.File = PROGRAM_TEMPORARY;
-		vpi[i].DstReg.Index = 0;
-		vpi[i].DstReg.WriteMask = 0xf;
-		vpi[i].DstReg.CondMask = COND_TR;
-
-		vpi[i].SrcReg[0].File = PROGRAM_STATE_VAR;
-		vpi[i].SrcReg[0].Index = idx;
-		vpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-
-		vpi[i].SrcReg[1].File = PROGRAM_INPUT;
-		vpi[i].SrcReg[1].Index = VERT_ATTRIB_POS;
-		vpi[i].SrcReg[1].Swizzle = MAKE_SWIZZLE4(i, i, i, i);
-
-		if (i > 0) {
-			vpi[i].SrcReg[2].File = PROGRAM_TEMPORARY;
-			vpi[i].SrcReg[2].Index = 0;
-			vpi[i].SrcReg[2].Swizzle = SWIZZLE_XYZW;
-		}
-#endif
-	}
-
-	_mesa_copy_instructions(&vpi[i], prog->Instructions,
-				prog->NumInstructions);
-
-	free(prog->Instructions);
-
-	prog->Instructions = vpi;
-
-	prog->NumInstructions += 4;
-	vpi = &prog->Instructions[prog->NumInstructions - 1];
-
-	assert(vpi->Opcode == OPCODE_END);
-}
-
 static void insert_wpos(struct r300_vertex_program *vp, struct gl_program *prog,
 			GLuint temp_index)
 {
@@ -1335,9 +1248,10 @@ static void pos_as_texcoord(struct r300_vertex_program *vp,
 	insert_wpos(vp, prog, tempregi);
 }
 
-static struct r300_vertex_program *build_program(struct r300_vertex_program_key
-						 *wanted_key, struct gl_vertex_program
-						 *mesa_vp, GLint wpos_idx)
+static struct r300_vertex_program *build_program(GLcontext *ctx,
+						 struct r300_vertex_program_key *wanted_key,
+						 struct gl_vertex_program *mesa_vp,
+						 GLint wpos_idx)
 {
 	struct r300_vertex_program *vp;
 
@@ -1346,7 +1260,7 @@ static struct r300_vertex_program *build_program(struct r300_vertex_program_key
 	vp->wpos_idx = wpos_idx;
 
 	if (mesa_vp->IsPositionInvariant) {
-		position_invariant(&mesa_vp->Base);
+		_mesa_insert_mvp_code(ctx, mesa_vp);
 	}
 
 	if (wpos_idx > -1) {
@@ -1480,7 +1394,7 @@ void r300SelectVertexShader(r300ContextPtr r300)
 		fflush(stdout);
 	}
 
-	vp = build_program(&wanted_key, &vpc->mesa_program, wpos_idx);
+	vp = build_program(ctx, &wanted_key, &vpc->mesa_program, wpos_idx);
 	vp->next = vpc->progs;
 	vpc->progs = vp;
 	r300->selected_vp = vp;
-- 
cgit v1.2.3


From 0b411a72f3cc3be7ecf9f4676d9860b2b56f084e Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Mon, 29 Jun 2009 21:52:39 +0200
Subject: r300: print vertex program after adding artificial output insts

---
 src/mesa/drivers/dri/r300/r300_vertprog.c | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index 5da9bda281..a0babeff92 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -1259,6 +1259,12 @@ static struct r300_vertex_program *build_program(GLcontext *ctx,
 	_mesa_memcpy(&vp->key, wanted_key, sizeof(vp->key));
 	vp->wpos_idx = wpos_idx;
 
+	if (RADEON_DEBUG & DEBUG_VERTS) {
+		fprintf(stderr, "Initial vertex program:\n");
+		_mesa_print_program(&mesa_vp->Base);
+		fflush(stdout);
+	}
+
 	if (mesa_vp->IsPositionInvariant) {
 		_mesa_insert_mvp_code(ctx, mesa_vp);
 	}
@@ -1267,12 +1273,6 @@ static struct r300_vertex_program *build_program(GLcontext *ctx,
 		pos_as_texcoord(vp, &mesa_vp->Base);
 	}
 
-	if (RADEON_DEBUG & DEBUG_VERTS) {
-		fprintf(stderr, "Vertex program after native rewrite:\n");
-		_mesa_print_program(&mesa_vp->Base);
-		fflush(stdout);
-	}
-
 	/* Some outputs may be artificially added, to match the inputs of the fragment program.
 	 * Issue 16 of vertex program spec says that all vertex attributes that are unwritten by
 	 * vertex program are undefined, so just use MOV [vertex_result], CONST[0]
@@ -1310,6 +1310,12 @@ static struct r300_vertex_program *build_program(GLcontext *ctx,
 		}
 	}
 
+	if (RADEON_DEBUG & DEBUG_VERTS) {
+		fprintf(stderr, "Vertex program after native rewrite:\n");
+		_mesa_print_program(&mesa_vp->Base);
+		fflush(stdout);
+	}
+
 	assert(mesa_vp->Base.NumInstructions);
 	vp->num_temporaries = mesa_vp->Base.NumTemporaries;
 	r300TranslateVertexShader(vp, mesa_vp->Base.Instructions);
@@ -1388,12 +1394,6 @@ void r300SelectVertexShader(r300ContextPtr r300)
 			return;
 		}
 
-	if (RADEON_DEBUG & DEBUG_VERTS) {
-		fprintf(stderr, "Initial vertex program:\n");
-		_mesa_print_program(&vpc->mesa_program.Base);
-		fflush(stdout);
-	}
-
 	vp = build_program(ctx, &wanted_key, &vpc->mesa_program, wpos_idx);
 	vp->next = vpc->progs;
 	vpc->progs = vp;
-- 
cgit v1.2.3


From 21db37d43245f97032fa21030279c3b47c901639 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Wed, 1 Jul 2009 18:43:14 +0200
Subject: r300: translate non native insts earlier for easier debugging

---
 src/mesa/drivers/dri/r300/r300_vertprog.c | 202 ++++++++++++++++++++++++++++++
 1 file changed, 202 insertions(+)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index a0babeff92..49fcb6ff2c 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -72,6 +72,33 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 		u_temp_i=VSF_MAX_FRAGMENT_TEMPS-1; \
 	} while (0)
 
+static GLuint combineSwizzles(GLuint src, GLuint swz_x, GLuint swz_y, GLuint swz_z, GLuint swz_w)
+{
+	GLuint ret = 0;
+
+	if (swz_x == SWIZZLE_ZERO || swz_x == SWIZZLE_ONE)
+		ret |= swz_x;
+	else
+		ret |= GET_SWZ(src, swz_x);
+
+	if (swz_y == SWIZZLE_ZERO || swz_y == SWIZZLE_ONE)
+		ret |= swz_y << 3;
+	else
+		ret |= GET_SWZ(src, swz_y) << 3;
+
+	if (swz_z == SWIZZLE_ZERO || swz_z == SWIZZLE_ONE)
+		ret |= swz_z << 6;
+	else
+		ret |= GET_SWZ(src, swz_z) << 6;
+
+	if (swz_w == SWIZZLE_ZERO || swz_w == SWIZZLE_ONE)
+		ret |= swz_w << 9;
+	else
+		ret |= GET_SWZ(src, swz_w) << 9;
+
+	return ret;
+}
+
 int r300VertexProgUpdateParams(GLcontext * ctx,
 			       struct r300_vertex_program_cont *vp, float *dst)
 {
@@ -1248,6 +1275,179 @@ static void pos_as_texcoord(struct r300_vertex_program *vp,
 	insert_wpos(vp, prog, tempregi);
 }
 
+static int translateABS(struct gl_program *prog, int pos)
+{
+	struct prog_instruction *inst;
+
+	inst = &prog->Instructions[pos];
+
+	inst->Opcode = OPCODE_MAX;
+	inst->SrcReg[1] = inst->SrcReg[0];
+	inst->SrcReg[1].Negate ^= NEGATE_XYZW;
+
+	return 0;
+}
+
+static int translateDP3(struct gl_program *prog, int pos)
+{
+	struct prog_instruction *inst;
+
+	inst = &prog->Instructions[pos];
+
+	inst->Opcode = OPCODE_DP4;
+	inst->SrcReg[0].Swizzle = combineSwizzles(inst->SrcReg[0].Swizzle, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
+
+	return 0;
+}
+
+static int translateDPH(struct gl_program *prog, int pos)
+{
+	struct prog_instruction *inst;
+
+	inst = &prog->Instructions[pos];
+
+	inst->Opcode = OPCODE_DP4;
+	inst->SrcReg[0].Swizzle = combineSwizzles(inst->SrcReg[0].Swizzle, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
+
+	return 0;
+}
+
+static int translateFLR(struct gl_program *prog, int pos)
+{
+	struct prog_instruction *inst;
+	struct prog_dst_register dst;
+	int tmp_idx;
+
+	tmp_idx = prog->NumTemporaries++;
+
+	_mesa_insert_instructions(prog, pos + 1, 1);
+
+	inst = &prog->Instructions[pos];
+	dst = inst->DstReg;
+
+	inst->Opcode = OPCODE_FRC;
+	inst->DstReg.File = PROGRAM_TEMPORARY;
+	inst->DstReg.Index = tmp_idx;
+	++inst;
+
+	inst->Opcode = OPCODE_ADD;
+	inst->DstReg = dst;
+	inst->SrcReg[0] = (inst-1)->SrcReg[0];
+	inst->SrcReg[1].File = PROGRAM_TEMPORARY;
+	inst->SrcReg[1].Index = tmp_idx;
+	inst->SrcReg[1].Negate = NEGATE_XYZW;
+
+	return 1;
+}
+
+static int translateSUB(struct gl_program *prog, int pos)
+{
+	struct prog_instruction *inst;
+
+	inst = &prog->Instructions[pos];
+
+	inst->Opcode = OPCODE_ADD;
+	inst->SrcReg[1].Negate ^= NEGATE_XYZW;
+
+	return 0;
+}
+
+static int translateSWZ(struct gl_program *prog, int pos)
+{
+	struct prog_instruction *inst;
+	GLuint orig_negate, orig_writemask;
+
+	inst = &prog->Instructions[pos];
+	orig_negate = inst->SrcReg[0].Negate;
+	orig_writemask = inst->DstReg.WriteMask;
+
+	inst->Opcode = OPCODE_MOV;
+
+	/* If all relevant components are either negated or not negated at the same time, we are ok.
+	 */
+	if ((orig_negate & orig_writemask) == 0 || (orig_negate & orig_writemask) == (NEGATE_XYZW & orig_writemask))
+		return 0;
+
+	_mesa_insert_instructions(prog, pos + 1, 1);
+
+	inst = &prog->Instructions[pos];
+	inst->DstReg.WriteMask = orig_writemask & (orig_negate ^ NEGATE_XYZW);
+	inst->SrcReg[0].Negate = NEGATE_NONE;
+	++inst;
+
+	*inst = *(inst-1);
+	inst->DstReg.WriteMask = orig_writemask & orig_negate;
+	inst->SrcReg[0].Negate = NEGATE_XYZW;
+
+	return 1;
+}
+
+static int translateXPD(struct gl_program *prog, int pos)
+{
+	struct prog_instruction *inst;
+	int tmp_idx;
+
+	tmp_idx = prog->NumTemporaries++;
+
+	_mesa_insert_instructions(prog, pos + 1, 1);
+
+	inst = &prog->Instructions[pos];
+
+	*(inst+1) = *inst;
+
+	inst->Opcode = OPCODE_MUL;
+	inst->DstReg.File = PROGRAM_TEMPORARY;
+	inst->DstReg.Index = tmp_idx;
+	inst->SrcReg[0].Swizzle = combineSwizzles(inst->SrcReg[0].Swizzle, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_W);
+	inst->SrcReg[1].Swizzle = combineSwizzles(inst->SrcReg[1].Swizzle, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W);
+	++inst;
+
+	inst->Opcode = OPCODE_MAD;
+	inst->SrcReg[0].Swizzle = combineSwizzles(inst->SrcReg[0].Swizzle, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W);
+	inst->SrcReg[1].Swizzle = combineSwizzles(inst->SrcReg[1].Swizzle, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_W);
+	inst->SrcReg[1].Negate ^= NEGATE_XYZW;
+	inst->SrcReg[2].File = PROGRAM_TEMPORARY;
+	inst->SrcReg[2].Index = tmp_idx;
+
+	return 1;
+}
+
+static void translateInsts(struct gl_program *prog)
+{
+	struct prog_instruction *inst;
+	int i;
+
+	for (i = 0; i < prog->NumInstructions; ++i) {
+		inst = &prog->Instructions[i];
+
+		switch (inst->Opcode) {
+			case OPCODE_ABS:
+				i += translateABS(prog, i);
+				break;
+			case OPCODE_DP3:
+				i += translateDP3(prog, i);
+				break;
+			case OPCODE_DPH:
+				i += translateDPH(prog, i);
+				break;
+			case OPCODE_FLR:
+				i += translateFLR(prog, i);
+				break;
+			case OPCODE_SUB:
+				i += translateSUB(prog, i);
+				break;
+			case OPCODE_SWZ:
+				i += translateSWZ(prog, i);
+				break;
+			case OPCODE_XPD:
+				i += translateXPD(prog, i);
+				break;
+			default:
+				break;
+		}
+	}
+}
+
 static struct r300_vertex_program *build_program(GLcontext *ctx,
 						 struct r300_vertex_program_key *wanted_key,
 						 struct gl_vertex_program *mesa_vp,
@@ -1310,6 +1510,8 @@ static struct r300_vertex_program *build_program(GLcontext *ctx,
 		}
 	}
 
+	translateInsts(&mesa_vp->Base);
+
 	if (RADEON_DEBUG & DEBUG_VERTS) {
 		fprintf(stderr, "Vertex program after native rewrite:\n");
 		_mesa_print_program(&mesa_vp->Base);
-- 
cgit v1.2.3


From 28066ed012b51f2171866669a2972bc7ee293565 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 4 Jul 2009 12:56:31 +0200
Subject: r300: update state parameters only once per rendering operation

---
 src/mesa/drivers/dri/r300/r300_fragprog_common.c | 11 -----------
 src/mesa/drivers/dri/r300/r300_state.c           |  6 +++---
 src/mesa/drivers/dri/r300/r300_state.h           |  1 -
 3 files changed, 3 insertions(+), 15 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.c b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
index abc8757ba1..1644e6f42a 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_common.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
@@ -50,13 +50,6 @@
 #include "radeon_program.h"
 #include "radeon_program_alu.h"
 
-static void update_params(GLcontext *ctx, struct gl_fragment_program *fp)
-{
-	/* Ask Mesa nicely to fill in ParameterValues for us */
-	if (fp->Base.Parameters)
-		_mesa_load_state_parameters(ctx, fp->Base.Parameters);
-}
-
 static void nqssadce_init(struct nqssadce_state* s)
 {
 	s->Outputs[FRAG_RESULT_COLOR].Sourced = WRITEMASK_XYZW;
@@ -281,11 +274,7 @@ void r300TranslateFragmentShader(GLcontext *ctx, struct gl_fragment_program *fp)
 
 		r300_fp->translated = GL_TRUE;
 
-		r300UpdateStateParameters(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
-
 		if (r300_fp->error || (RADEON_DEBUG & DEBUG_PIXEL))
 			r300->vtbl.FragmentProgramDump(&r300_fp->code);
 	}
-
-	update_params(ctx, fp);
 }
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index c0eda977db..67bd6ec33b 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -1093,7 +1093,7 @@ r300FetchStateParameter(GLcontext * ctx,
  * Update R300's own internal state parameters.
  * For now just STATE_R300_WINDOW_DIMENSION
  */
-void r300UpdateStateParameters(GLcontext * ctx, GLuint new_state)
+static void r300UpdateStateParameters(GLcontext * ctx, GLuint new_state)
 {
 	struct r300_fragment_program *fp;
 	struct gl_program_parameter_list *paramList;
@@ -1111,6 +1111,8 @@ void r300UpdateStateParameters(GLcontext * ctx, GLuint new_state)
 	if (!paramList)
 		return;
 
+	_mesa_load_state_parameters(ctx, paramList);
+
 	for (i = 0; i < paramList->NumParameters; i++) {
 		if (paramList->Parameters[i].Type == PROGRAM_STATE_VAR) {
 			r300FetchStateParameter(ctx,
@@ -2323,8 +2325,6 @@ static void r300InvalidateState(GLcontext * ctx, GLuint new_state)
 		R300_STATECHANGE(r300, cb);
 	}
 
-	r300UpdateStateParameters(ctx, new_state);
-
 	r300->radeon.NewGLState |= new_state;
 }
 
diff --git a/src/mesa/drivers/dri/r300/r300_state.h b/src/mesa/drivers/dri/r300/r300_state.h
index 2328289420..d46bf9f179 100644
--- a/src/mesa/drivers/dri/r300/r300_state.h
+++ b/src/mesa/drivers/dri/r300/r300_state.h
@@ -52,7 +52,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 void r300UpdateViewportOffset (GLcontext * ctx);
 void r300UpdateDrawBuffer (GLcontext * ctx);
-void r300UpdateStateParameters (GLcontext * ctx, GLuint new_state);
 void r300UpdateShaders (r300ContextPtr rmesa);
 void r300UpdateShaderStates (r300ContextPtr rmesa);
 void r300InitState (r300ContextPtr r300);
-- 
cgit v1.2.3


From 7829b7a1b85dd8e6c31189e7f3dec91d71d134c3 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 4 Jul 2009 13:26:49 +0200
Subject: r300: cache translated fragment programs

---
 src/mesa/drivers/dri/r300/r300_context.h         |  10 +-
 src/mesa/drivers/dri/r300/r300_fragprog_common.c | 169 ++++++++++++-----------
 src/mesa/drivers/dri/r300/r300_fragprog_common.h |   6 +-
 src/mesa/drivers/dri/r300/r300_shader.c          |  51 ++++++-
 src/mesa/drivers/dri/r300/r300_state.c           |  59 ++++----
 src/mesa/drivers/dri/r300/r300_swtcl.c           |   2 +-
 6 files changed, 173 insertions(+), 124 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 026c33c67c..fc436e1100 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -546,7 +546,7 @@ struct r500_fragment_program_code {
 * to render with that program.
 */
 struct r300_fragment_program {
-	struct gl_fragment_program Base;
+	struct gl_program *Base;
 
 	GLboolean translated;
 	GLboolean error;
@@ -559,6 +559,13 @@ struct r300_fragment_program {
 
 	GLboolean writes_depth;
 	GLuint optimization;
+
+	struct r300_fragment_program *next;
+};
+
+struct r300_fragment_program_cont {
+	struct gl_fragment_program Base;
+	struct r300_fragment_program *progs;
 };
 
 struct r300_fragment_program_compiler {
@@ -633,6 +640,7 @@ struct r300_context {
 	struct r300_hw_state hw;
 
 	struct r300_vertex_program *selected_vp;
+	struct r300_fragment_program *selected_fp;
 
 	/* Vertex buffers
 	 */
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.c b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
index 1644e6f42a..b25cf24007 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_common.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
@@ -67,7 +67,7 @@ static void nqssadce_init(struct nqssadce_state* s)
  */
 static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler)
 {
-	GLuint InputsRead = compiler->fp->Base.Base.InputsRead;
+	GLuint InputsRead = compiler->fp->Base->InputsRead;
 
 	if (!(InputsRead & FRAG_BIT_WPOS))
 		return;
@@ -168,7 +168,7 @@ static GLuint build_func(GLuint comparefunc)
  */
 static void build_state(
 	r300ContextPtr r300,
-	struct r300_fragment_program *fp,
+	struct gl_fragment_program *fp,
 	struct r300_fragment_program_external_state *state)
 {
 	int unit;
@@ -176,7 +176,7 @@ static void build_state(
 	_mesa_bzero(state, sizeof(*state));
 
 	for(unit = 0; unit < 16; ++unit) {
-		if (fp->Base.Base.ShadowSamplers & (1 << unit)) {
+		if (fp->Base.ShadowSamplers & (1 << unit)) {
 			struct gl_texture_object* tex = r300->radeon.glCtx->Texture.Unit[unit]._Current;
 
 			state->unit[unit].depth_texture_mode = build_dtm(tex->DepthMode);
@@ -185,96 +185,107 @@ static void build_state(
 	}
 }
 
-void r300TranslateFragmentShader(GLcontext *ctx, struct gl_fragment_program *fp)
+void r300TranslateFragmentShader(GLcontext *ctx, struct r300_fragment_program *fp)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-	struct r300_fragment_program *r300_fp = (struct r300_fragment_program *)fp;
-	struct r300_fragment_program_external_state state;
+	struct r300_fragment_program_compiler compiler;
+
+	compiler.r300 = r300;
+	compiler.fp = fp;
+	compiler.code = &fp->code;
+	compiler.program = fp->Base;
+
+	if (RADEON_DEBUG & DEBUG_PIXEL) {
+		fflush(stdout);
+		_mesa_printf("Fragment Program: Initial program:\n");
+		_mesa_print_program(compiler.program);
+		fflush(stdout);
+	}
 
-	build_state(r300, r300_fp, &state);
-	if (_mesa_memcmp(&r300_fp->state, &state, sizeof(state))) {
-		/* TODO: cache compiled programs */
-		r300_fp->translated = GL_FALSE;
-		_mesa_memcpy(&r300_fp->state, &state, sizeof(state));
+	insert_WPOS_trailer(&compiler);
+
+	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+		struct radeon_program_transformation transformations[] = {
+			{ &r500_transform_TEX, &compiler },
+			{ &radeonTransformALU, 0 },
+			{ &radeonTransformDeriv, 0 },
+			{ &radeonTransformTrigScale, 0 }
+		};
+		radeonLocalTransform(ctx, compiler.program, 4, transformations);
+	} else {
+		struct radeon_program_transformation transformations[] = {
+			{ &r300_transform_TEX, &compiler },
+			{ &radeonTransformALU, 0 },
+			{ &radeonTransformTrigSimple, 0 }
+		};
+		radeonLocalTransform(ctx, compiler.program, 3, transformations);
 	}
 
-	if (!r300_fp->translated) {
-		struct r300_fragment_program_compiler compiler;
+	if (RADEON_DEBUG & DEBUG_PIXEL) {
+		_mesa_printf("Fragment Program: After native rewrite:\n");
+		_mesa_print_program(compiler.program);
+		fflush(stdout);
+	}
 
-		compiler.r300 = r300;
-		compiler.fp = r300_fp;
-		compiler.code = &r300_fp->code;
-		compiler.program = _mesa_clone_program(ctx, &fp->Base);
+	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+		struct radeon_nqssadce_descr nqssadce = {
+			.Init = &nqssadce_init,
+			.IsNativeSwizzle = &r500FPIsNativeSwizzle,
+			.BuildSwizzle = &r500FPBuildSwizzle,
+			.RewriteDepthOut = GL_TRUE
+		};
+		radeonNqssaDce(ctx, compiler.program, &nqssadce);
+	} else {
+		struct radeon_nqssadce_descr nqssadce = {
+			.Init = &nqssadce_init,
+			.IsNativeSwizzle = &r300FPIsNativeSwizzle,
+			.BuildSwizzle = &r300FPBuildSwizzle,
+			.RewriteDepthOut = GL_TRUE
+		};
+		radeonNqssaDce(ctx, compiler.program, &nqssadce);
+	}
 
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			fflush(stdout);
-			_mesa_printf("Fragment Program: Initial program:\n");
-			_mesa_print_program(compiler.program);
-			fflush(stdout);
-		}
+	if (RADEON_DEBUG & DEBUG_PIXEL) {
+		_mesa_printf("Compiler: after NqSSA-DCE:\n");
+		_mesa_print_program(compiler.program);
+		fflush(stdout);
+	}
 
-		insert_WPOS_trailer(&compiler);
-
-		if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
-			struct radeon_program_transformation transformations[] = {
-				{ &r500_transform_TEX, &compiler },
-				{ &radeonTransformALU, 0 },
-				{ &radeonTransformDeriv, 0 },
-				{ &radeonTransformTrigScale, 0 }
-			};
-			radeonLocalTransform(ctx, compiler.program, 4, transformations);
-		} else {
-			struct radeon_program_transformation transformations[] = {
-				{ &r300_transform_TEX, &compiler },
-				{ &radeonTransformALU, 0 },
-				{ &radeonTransformTrigSimple, 0 }
-			};
-			radeonLocalTransform(ctx, compiler.program, 3, transformations);
-		}
+	if (!r300->vtbl.BuildFragmentProgramHwCode(&compiler))
+		fp->error = GL_TRUE;
 
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Fragment Program: After native rewrite:\n");
-			_mesa_print_program(compiler.program);
-			fflush(stdout);
-		}
+	fp->translated = GL_TRUE;
 
-		if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
-			struct radeon_nqssadce_descr nqssadce = {
-				.Init = &nqssadce_init,
-				.IsNativeSwizzle = &r500FPIsNativeSwizzle,
-				.BuildSwizzle = &r500FPBuildSwizzle,
-				.RewriteDepthOut = GL_TRUE
-			};
-			radeonNqssaDce(ctx, compiler.program, &nqssadce);
-		} else {
-			struct radeon_nqssadce_descr nqssadce = {
-				.Init = &nqssadce_init,
-				.IsNativeSwizzle = &r300FPIsNativeSwizzle,
-				.BuildSwizzle = &r300FPBuildSwizzle,
-				.RewriteDepthOut = GL_TRUE
-			};
-			radeonNqssaDce(ctx, compiler.program, &nqssadce);
-		}
+	if (fp->error || (RADEON_DEBUG & DEBUG_PIXEL))
+		r300->vtbl.FragmentProgramDump(&fp->code);
+}
 
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Compiler: after NqSSA-DCE:\n");
-			_mesa_print_program(compiler.program);
-			fflush(stdout);
-		}
+struct r300_fragment_program *r300SelectFragmentShader(GLcontext *ctx)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_fragment_program_cont *fp_list;
+	struct r300_fragment_program *fp;
+	struct r300_fragment_program_external_state state;
+
+	fp_list = (struct r300_fragment_program_cont *)ctx->FragmentProgram._Current;
+	build_state(r300, ctx->FragmentProgram._Current, &state);
 
-		if (!r300->vtbl.BuildFragmentProgramHwCode(&compiler))
-			r300_fp->error = GL_TRUE;
+	fp = fp_list->progs;
+	while (fp) {
+	       if (_mesa_memcmp(&fp->state, &state, sizeof(state)) == 0) {
+	               return r300->selected_fp = fp;
+	       }
+	       fp = fp->next;
+	}
 
-		/* Subtle: Rescue any parameters that have been added during transformations */
-		_mesa_free_parameter_list(fp->Base.Parameters);
-		fp->Base.Parameters = compiler.program->Parameters;
-		compiler.program->Parameters = 0;
+	fp = _mesa_calloc(sizeof(struct r300_fragment_program));
 
-		_mesa_reference_program(ctx, &compiler.program, NULL);
+	fp->state = state;
+	fp->translated = GL_FALSE;
+	fp->Base = _mesa_clone_program(ctx, &ctx->FragmentProgram._Current->Base);
 
-		r300_fp->translated = GL_TRUE;
+	fp->next = fp_list->progs;
+	fp_list->progs = fp;
 
-		if (r300_fp->error || (RADEON_DEBUG & DEBUG_PIXEL))
-			r300->vtbl.FragmentProgramDump(&r300_fp->code);
-	}
+	return r300->selected_fp = fp;
 }
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.h b/src/mesa/drivers/dri/r300/r300_fragprog_common.h
index 85ea86fecb..5e103ee408 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_common.h
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.h
@@ -30,6 +30,10 @@
 
 #include "main/mtypes.h"
 
-extern void r300TranslateFragmentShader(GLcontext *ctx, struct gl_fragment_program *fp);
+#include "r300_context.h"
+
+extern void r300TranslateFragmentShader(GLcontext *ctx, struct r300_fragment_program *fp);
+
+struct r300_fragment_program *r300SelectFragmentShader(GLcontext *ctx);
 
 #endif
diff --git a/src/mesa/drivers/dri/r300/r300_shader.c b/src/mesa/drivers/dri/r300/r300_shader.c
index 0133b83796..c9a429943a 100644
--- a/src/mesa/drivers/dri/r300/r300_shader.c
+++ b/src/mesa/drivers/dri/r300/r300_shader.c
@@ -32,11 +32,34 @@
 #include "r300_context.h"
 #include "r300_fragprog_common.h"
 
+static void freeFragProgCache(GLcontext *ctx, struct r300_fragment_program_cont *cache)
+{
+	struct r300_fragment_program *tmp, *fp = cache->progs;
+
+	while (fp) {
+		tmp = fp->next;
+		_mesa_reference_program(ctx, &fp->Base, NULL);
+		_mesa_free(fp);
+		fp = tmp;
+	}
+}
+
+static void freeVertProgCache(GLcontext *ctx, struct r300_vertex_program_cont *cache)
+{
+	struct r300_vertex_program *tmp, *vp = cache->progs;
+
+	while (vp) {
+		tmp = vp->next;
+		_mesa_free(vp);
+		vp = tmp;
+	}
+}
+
 static struct gl_program *r300NewProgram(GLcontext * ctx, GLenum target,
 					 GLuint id)
 {
 	struct r300_vertex_program_cont *vp;
-	struct r300_fragment_program *fp;
+	struct r300_fragment_program_cont *fp;
 
 	switch (target) {
 	case GL_VERTEX_STATE_PROGRAM_NV:
@@ -47,7 +70,7 @@ static struct gl_program *r300NewProgram(GLcontext * ctx, GLenum target,
 
 	case GL_FRAGMENT_PROGRAM_NV:
 	case GL_FRAGMENT_PROGRAM_ARB:
-		fp = CALLOC_STRUCT(r300_fragment_program);
+		fp = CALLOC_STRUCT(r300_fragment_program_cont);
 		return _mesa_init_fragment_program(ctx, &fp->Base, target, id);
 
 	default:
@@ -59,21 +82,35 @@ static struct gl_program *r300NewProgram(GLcontext * ctx, GLenum target,
 
 static void r300DeleteProgram(GLcontext * ctx, struct gl_program *prog)
 {
+	struct r300_vertex_program_cont *vp = (struct r300_vertex_program_cont *)prog;
+	struct r300_fragment_program_cont *fp = (struct r300_fragment_program_cont *)prog;
+
+	switch (prog->Target) {
+		case GL_VERTEX_PROGRAM_ARB:
+			freeVertProgCache(ctx, vp);
+			break;
+		case GL_FRAGMENT_PROGRAM_ARB:
+			freeFragProgCache(ctx, fp);
+			break;
+	}
+
 	_mesa_delete_program(ctx, prog);
 }
 
 static void
 r300ProgramStringNotify(GLcontext * ctx, GLenum target, struct gl_program *prog)
 {
-	struct r300_vertex_program_cont *vp = (void *)prog;
-	struct r300_fragment_program *r300_fp = (struct r300_fragment_program *)prog;
+	struct r300_vertex_program_cont *vp = (struct r300_vertex_program_cont *)prog;
+	struct r300_fragment_program_cont *fp = (struct r300_fragment_program_cont *)prog;
 
 	switch (target) {
 	case GL_VERTEX_PROGRAM_ARB:
+		freeVertProgCache(ctx, vp);
 		vp->progs = NULL;
 		break;
 	case GL_FRAGMENT_PROGRAM_ARB:
-		r300_fp->translated = GL_FALSE;
+		freeFragProgCache(ctx, fp);
+		fp->progs = NULL;
 		break;
 	}
 
@@ -85,9 +122,9 @@ static GLboolean
 r300IsProgramNative(GLcontext * ctx, GLenum target, struct gl_program *prog)
 {
 	if (target == GL_FRAGMENT_PROGRAM_ARB) {
-		struct r300_fragment_program *fp = (struct r300_fragment_program *)prog;
+		struct r300_fragment_program *fp = r300SelectFragmentShader(ctx);
 		if (!fp->translated)
-			r300TranslateFragmentShader(ctx, &fp->Base);
+			r300TranslateFragmentShader(ctx, fp);
 
 		return !fp->error;
 	} else
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index 67bd6ec33b..e0996383eb 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -452,9 +452,9 @@ static void r300SetPolygonOffsetState(GLcontext * ctx, GLboolean state)
 
 static GLboolean current_fragment_program_writes_depth(GLcontext* ctx)
 {
-	struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current;
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
 
-	return (fp && fp->writes_depth);
+	return ctx->FragmentProgram._Current && r300->selected_fp->writes_depth;
 }
 
 static void r300SetEarlyZState(GLcontext * ctx)
@@ -1095,18 +1095,17 @@ r300FetchStateParameter(GLcontext * ctx,
  */
 static void r300UpdateStateParameters(GLcontext * ctx, GLuint new_state)
 {
-	struct r300_fragment_program *fp;
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct gl_program_parameter_list *paramList;
 	GLuint i;
 
 	if (!(new_state & (_NEW_BUFFERS | _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS)))
 		return;
 
-	fp = (struct r300_fragment_program *)ctx->FragmentProgram._Current;
-	if (!fp)
+	if (!ctx->FragmentProgram._Current || !rmesa->selected_fp)
 		return;
 
-	paramList = fp->Base.Base.Parameters;
+	paramList = rmesa->selected_fp->Base->Parameters;
 
 	if (!paramList)
 		return;
@@ -1227,8 +1226,7 @@ static void r300SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	int i;
-	struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current;
-	struct r300_fragment_program_code *code = &fp->code.r300;
+	struct r300_fragment_program_code *code = &r300->selected_fp->code.r300;
 
 	R300_STATECHANGE(r300, fpt);
 
@@ -1268,9 +1266,9 @@ static void r300SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings)
 
 static void r500SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings)
 {
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	int i;
-	struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current;
-	struct r500_fragment_program_code *code = &fp->code.r500;
+	struct r500_fragment_program_code *code = &r300->selected_fp->code.r500;
 
 	/* find all the texture instructions and relocate the texture units */
 	for (i = 0; i < code->inst_end + 1; i++) {
@@ -1318,8 +1316,6 @@ static void r300SetupTextures(GLcontext * ctx)
 	int hw_tmu = 0;
 	int last_hw_tmu = -1;	/* -1 translates into no setup costs for fields */
 	int tmu_mappings[R300_MAX_TEXTURE_UNITS] = { -1, };
-	struct r300_fragment_program *fp = (struct r300_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
 
 	R300_STATECHANGE(r300, txe);
 	R300_STATECHANGE(r300, tex.filter);
@@ -1422,7 +1418,7 @@ static void r300SetupTextures(GLcontext * ctx)
 	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_BORDER_COLOR_0, last_hw_tmu + 1);
 
 	if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) {
-		if (fp->Base.UsesKill && last_hw_tmu < 0) {
+		if (ctx->FragmentProgram._Current->UsesKill && last_hw_tmu < 0) {
 			// The KILL operation requires the first texture unit
 			// to be enabled.
 			r300->hw.txe.cmd[R300_TXE_ENABLE] |= 1;
@@ -1464,7 +1460,7 @@ static void r300SetupRSUnit(GLcontext * ctx)
 	else
 		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->render_inputs_bitset);
 
-	InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
+	InputsRead = r300->selected_fp->Base->InputsRead;
 
 	R300_STATECHANGE(r300, ri);
 	R300_STATECHANGE(r300, rc);
@@ -1581,7 +1577,7 @@ static void r500SetupRSUnit(GLcontext * ctx)
 	else
 		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->render_inputs_bitset);
 
-	InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
+	InputsRead = r300->selected_fp->Base->InputsRead;
 
 	R300_STATECHANGE(r300, ri);
 	R300_STATECHANGE(r300, rc);
@@ -2038,21 +2034,18 @@ static void r300ResetHwState(r300ContextPtr r300)
 
 void r300UpdateShaders(r300ContextPtr rmesa)
 {
-	GLcontext *ctx;
+	GLcontext *ctx = rmesa->radeon.glCtx;
 	struct r300_fragment_program *fp;
-	int i;
-
-	ctx = rmesa->radeon.glCtx;
-	fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current;
 
 	/* should only happenen once, just after context is created */
 	/* TODO: shouldn't we fallback to sw here? */
-	if (!fp) {
+	if (!ctx->FragmentProgram._Current) {
 		_mesa_fprintf(stderr, "No ctx->FragmentProgram._Current!!\n");
 		return;
 	}
 
 	if (rmesa->radeon.NewGLState && rmesa->options.hw_tcl_enabled) {
+		int i;
 		for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
 			rmesa->temp_attrib[i] =
 			    TNL_CONTEXT(ctx)->vb.AttribPtr[i];
@@ -2071,8 +2064,9 @@ void r300UpdateShaders(r300ContextPtr rmesa)
 		r300SwitchFallback(ctx, R300_FALLBACK_VERTEX_PROGRAM, rmesa->selected_vp->error);
 	}
 
-	if (!fp->translated || rmesa->radeon.NewGLState)
-		r300TranslateFragmentShader(ctx, ctx->FragmentProgram._Current);
+	fp = r300SelectFragmentShader(ctx);
+	if (!fp->translated)
+		r300TranslateFragmentShader(ctx, fp);
 
 	r300SwitchFallback(ctx, R300_FALLBACK_FRAGMENT_PROGRAM, fp->error);
 
@@ -2104,7 +2098,7 @@ static const GLfloat *get_fragmentprogram_constant(GLcontext *ctx,
 static void r300SetupPixelShader(GLcontext *ctx)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current;
+	struct r300_fragment_program *fp = rmesa->selected_fp;
 	struct r300_fragment_program_code *code;
 	int i, k;
 
@@ -2150,8 +2144,7 @@ static void r300SetupPixelShader(GLcontext *ctx)
 	R300_STATECHANGE(rmesa, fpp);
 	rmesa->hw.fpp.cmd[R300_FPP_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_PFS_PARAM_0_X, code->const_nr * 4);
 	for (i = 0; i < code->const_nr; i++) {
-		const GLfloat *constant = get_fragmentprogram_constant(ctx,
-			&fp->Base.Base, code->constant[i]);
+		const GLfloat *constant = get_fragmentprogram_constant(ctx, fp->Base, code->constant[i]);
 		rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 0] = r300PackFloat24(constant[0]);
 		rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 1] = r300PackFloat24(constant[1]);
 		rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 2] = r300PackFloat24(constant[2]);
@@ -2176,7 +2169,7 @@ static void r300SetupPixelShader(GLcontext *ctx)
 static void r500SetupPixelShader(GLcontext *ctx)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current;
+	struct r300_fragment_program *fp = rmesa->selected_fp;
 	int i;
 	struct r500_fragment_program_code *code;
 
@@ -2212,8 +2205,7 @@ static void r500SetupPixelShader(GLcontext *ctx)
 
 	R300_STATECHANGE(rmesa, r500fp_const);
 	for (i = 0; i < code->const_nr; i++) {
-		const GLfloat *constant = get_fragmentprogram_constant(ctx,
-			&fp->Base.Base, code->constant[i]);
+		const GLfloat *constant = get_fragmentprogram_constant(ctx, fp->Base, code->constant[i]);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 0] = r300PackFloat32(constant[0]);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 1] = r300PackFloat32(constant[1]);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 2] = r300PackFloat32(constant[2]);
@@ -2276,20 +2268,17 @@ void r300SetupVAP(GLcontext *ctx, GLuint InputsRead, GLuint OutputsWritten)
 
 	rmesa->hw.vic.cmd[R300_VIC_CNTL_0] = r300VAPInputCntl0(ctx, InputsRead);
 	rmesa->hw.vic.cmd[R300_VIC_CNTL_1] = r300VAPInputCntl1(ctx, InputsRead);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_0] = r300VAPOutputCntl0(ctx, OutputsWritten, ctx->FragmentProgram._Current->Base.InputsRead);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = r300VAPOutputCntl1(ctx, OutputsWritten, ctx->FragmentProgram._Current->Base.InputsRead);
+	rmesa->hw.vof.cmd[R300_VOF_CNTL_0] = r300VAPOutputCntl0(ctx, OutputsWritten, rmesa->selected_fp->Base->InputsRead);
+	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = r300VAPOutputCntl1(ctx, OutputsWritten, rmesa->selected_fp->Base->InputsRead);
 }
 
 void r300UpdateShaderStates(r300ContextPtr rmesa)
 {
 	GLcontext *ctx;
 	ctx = rmesa->radeon.glCtx;
-	struct r300_fragment_program *r300_fp;
-
-	r300_fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current;
 
 	/* should only happenen once, just after context is created */
-	if (!r300_fp)
+	if (!ctx->FragmentProgram._Current)
 		return;
 
 	r300SetEarlyZState(ctx);
diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c
index ce4179208e..db4ccce6f1 100644
--- a/src/mesa/drivers/dri/r300/r300_swtcl.c
+++ b/src/mesa/drivers/dri/r300/r300_swtcl.c
@@ -76,7 +76,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead,  GLuint *_
 	GLuint InputsRead = 0;
 	GLuint OutputsWritten = 0;
 	int num_attrs = 0;
-	GLuint fp_reads = ctx->FragmentProgram._Current->Base.InputsRead;
+	GLuint fp_reads = rmesa->selected_fp->Base->InputsRead;
 	struct vertex_attribute *attrs = rmesa->vbuf.attribs;
 
 	rmesa->swtcl.coloroffset = rmesa->swtcl.specoffset = 0;
-- 
cgit v1.2.3


From bce224c1f108e6c8131dfc953ef607689b83ae7e Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Fri, 3 Jul 2009 20:06:23 +0200
Subject: r300: don't modify original vertex program

Keep the original vertex program untouched because it may be needed after some state change for generating new r300 specific vertex program.
---
 src/mesa/drivers/dri/r300/r300_context.h  |  4 +--
 src/mesa/drivers/dri/r300/r300_shader.c   |  1 +
 src/mesa/drivers/dri/r300/r300_vertprog.c | 51 ++++++++++++++++---------------
 3 files changed, 28 insertions(+), 28 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index fc436e1100..30b6e75039 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -405,6 +405,7 @@ struct r300_hw_state {
 #undef TAG
 
 struct r300_vertex_program {
+	struct gl_vertex_program *Base;
 	struct r300_vertex_program *next;
 
 	struct r300_vertex_program_key {
@@ -674,9 +675,6 @@ extern GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 
 extern void r300SelectVertexShader(r300ContextPtr r300);
 extern void r300InitShaderFuncs(struct dd_function_table *functions);
-extern int r300VertexProgUpdateParams(GLcontext * ctx,
-				      struct r300_vertex_program_cont *vp,
-				      float *dst);
 
 extern void r300InitShaderFunctions(r300ContextPtr r300);
 
diff --git a/src/mesa/drivers/dri/r300/r300_shader.c b/src/mesa/drivers/dri/r300/r300_shader.c
index c9a429943a..7206379ea0 100644
--- a/src/mesa/drivers/dri/r300/r300_shader.c
+++ b/src/mesa/drivers/dri/r300/r300_shader.c
@@ -50,6 +50,7 @@ static void freeVertProgCache(GLcontext *ctx, struct r300_vertex_program_cont *c
 
 	while (vp) {
 		tmp = vp->next;
+		_mesa_reference_vertprog(ctx, &vp->Base, NULL);
 		_mesa_free(vp);
 		vp = tmp;
 	}
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index 49fcb6ff2c..25c42814ec 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -99,15 +99,13 @@ static GLuint combineSwizzles(GLuint src, GLuint swz_x, GLuint swz_y, GLuint swz
 	return ret;
 }
 
-int r300VertexProgUpdateParams(GLcontext * ctx,
-			       struct r300_vertex_program_cont *vp, float *dst)
+static int r300VertexProgUpdateParams(GLcontext * ctx, struct gl_vertex_program *vp, float *dst)
 {
 	int pi;
-	struct gl_vertex_program *mesa_vp = &vp->mesa_program;
 	float *dst_o = dst;
 	struct gl_program_parameter_list *paramList;
 
-	if (mesa_vp->IsNVProgram) {
+	if (vp->IsNVProgram) {
 		_mesa_load_tracked_matrices(ctx);
 
 		for (pi = 0; pi < MAX_NV_VERTEX_PROGRAM_PARAMS; pi++) {
@@ -119,16 +117,18 @@ int r300VertexProgUpdateParams(GLcontext * ctx,
 		return dst - dst_o;
 	}
 
-	assert(mesa_vp->Base.Parameters);
-	_mesa_load_state_parameters(ctx, mesa_vp->Base.Parameters);
+	if (!vp->Base.Parameters)
+		return 0;
+
+	_mesa_load_state_parameters(ctx, vp->Base.Parameters);
 
-	if (mesa_vp->Base.Parameters->NumParameters * 4 >
+	if (vp->Base.Parameters->NumParameters * 4 >
 	    VSF_MAX_FRAGMENT_LENGTH) {
 		fprintf(stderr, "%s:Params exhausted\n", __FUNCTION__);
 		_mesa_exit(-1);
 	}
 
-	paramList = mesa_vp->Base.Parameters;
+	paramList = vp->Base.Parameters;
 	for (pi = 0; pi < paramList->NumParameters; pi++) {
 		switch (paramList->Parameters[pi].Type) {
 		case PROGRAM_STATE_VAR:
@@ -1027,9 +1027,9 @@ static void t_inputs_outputs(struct r300_vertex_program *vp)
 	}
 }
 
-static void r300TranslateVertexShader(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi)
+static void r300TranslateVertexShader(struct r300_vertex_program *vp)
 {
+	struct prog_instruction *vpi = vp->Base->Base.Instructions;
 	int i;
 	GLuint *inst;
 	unsigned long num_operands;
@@ -1450,27 +1450,31 @@ static void translateInsts(struct gl_program *prog)
 
 static struct r300_vertex_program *build_program(GLcontext *ctx,
 						 struct r300_vertex_program_key *wanted_key,
-						 struct gl_vertex_program *mesa_vp,
+						 const struct gl_vertex_program *mesa_vp,
 						 GLint wpos_idx)
 {
 	struct r300_vertex_program *vp;
+	struct gl_program *prog;
 
 	vp = _mesa_calloc(sizeof(*vp));
+	vp->Base = (struct gl_vertex_program *) _mesa_clone_program(ctx, &mesa_vp->Base);
 	_mesa_memcpy(&vp->key, wanted_key, sizeof(vp->key));
 	vp->wpos_idx = wpos_idx;
 
+	prog = &vp->Base->Base;
+
 	if (RADEON_DEBUG & DEBUG_VERTS) {
 		fprintf(stderr, "Initial vertex program:\n");
-		_mesa_print_program(&mesa_vp->Base);
+		_mesa_print_program(prog);
 		fflush(stdout);
 	}
 
 	if (mesa_vp->IsPositionInvariant) {
-		_mesa_insert_mvp_code(ctx, mesa_vp);
+		_mesa_insert_mvp_code(ctx, vp->Base);
 	}
 
 	if (wpos_idx > -1) {
-		pos_as_texcoord(vp, &mesa_vp->Base);
+		pos_as_texcoord(vp, prog);
 	}
 
 	/* Some outputs may be artificially added, to match the inputs of the fragment program.
@@ -1488,8 +1492,8 @@ static struct r300_vertex_program *build_program(GLcontext *ctx,
 		if (count > 0) {
 			struct prog_instruction *inst;
 
-			_mesa_insert_instructions(&mesa_vp->Base, mesa_vp->Base.NumInstructions - 1, count);
-			inst = &mesa_vp->Base.Instructions[mesa_vp->Base.NumInstructions - 1 - count];
+			_mesa_insert_instructions(prog, prog->NumInstructions - 1, count);
+			inst = &prog->Instructions[prog->NumInstructions - 1 - count];
 
 			for (i = 0; i < VERT_RESULT_MAX; ++i) {
 				if (vp->key.OutputsAdded & (1 << i)) {
@@ -1510,17 +1514,17 @@ static struct r300_vertex_program *build_program(GLcontext *ctx,
 		}
 	}
 
-	translateInsts(&mesa_vp->Base);
+	translateInsts(prog);
 
 	if (RADEON_DEBUG & DEBUG_VERTS) {
 		fprintf(stderr, "Vertex program after native rewrite:\n");
-		_mesa_print_program(&mesa_vp->Base);
+		_mesa_print_program(prog);
 		fflush(stdout);
 	}
 
-	assert(mesa_vp->Base.NumInstructions);
-	vp->num_temporaries = mesa_vp->Base.NumTemporaries;
-	r300TranslateVertexShader(vp, mesa_vp->Base.Instructions);
+	assert(prog->NumInstructions);
+	vp->num_temporaries = prog->NumTemporaries;
+	r300TranslateVertexShader(vp);
 
 	return vp;
 }
@@ -1653,10 +1657,7 @@ void r300SetupVertexProgram(r300ContextPtr rmesa)
 	((drm_r300_cmd_header_t *) rmesa->hw.vps.cmd)->vpu.count = 0;
 	
 	R300_STATECHANGE(rmesa, vpp);
-	param_count = r300VertexProgUpdateParams(ctx,
-								(struct r300_vertex_program_cont *)
-								ctx->VertexProgram._Current,
-								(float *)&rmesa->hw.vpp.cmd[R300_VPP_PARAM_0]);
+	param_count = r300VertexProgUpdateParams(ctx, prog->Base, (float *)&rmesa->hw.vpp.cmd[R300_VPP_PARAM_0]);
 	bump_vpu_count(rmesa->hw.vpp.cmd, param_count);
 	param_count /= 4;
 
-- 
cgit v1.2.3


From 37c319f62f59d7750dd172034e43dfd489f572cc Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Fri, 3 Jul 2009 20:14:24 +0200
Subject: r300: implement proper IsProgramNative check for vertex programs

---
 src/mesa/drivers/dri/r300/r300_context.h  |  1 -
 src/mesa/drivers/dri/r300/r300_shader.c   |  9 +++++++--
 src/mesa/drivers/dri/r300/r300_state.c    |  8 ++++++--
 src/mesa/drivers/dri/r300/r300_vertprog.c | 13 ++++++-------
 src/mesa/drivers/dri/r300/r300_vertprog.h |  4 ++++
 5 files changed, 23 insertions(+), 12 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 30b6e75039..44211a45b3 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -673,7 +673,6 @@ extern GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 				   __DRIcontextPrivate * driContextPriv,
 				   void *sharedContextPrivate);
 
-extern void r300SelectVertexShader(r300ContextPtr r300);
 extern void r300InitShaderFuncs(struct dd_function_table *functions);
 
 extern void r300InitShaderFunctions(r300ContextPtr r300);
diff --git a/src/mesa/drivers/dri/r300/r300_shader.c b/src/mesa/drivers/dri/r300/r300_shader.c
index 7206379ea0..854eb5d80a 100644
--- a/src/mesa/drivers/dri/r300/r300_shader.c
+++ b/src/mesa/drivers/dri/r300/r300_shader.c
@@ -128,8 +128,13 @@ r300IsProgramNative(GLcontext * ctx, GLenum target, struct gl_program *prog)
 			r300TranslateFragmentShader(ctx, fp);
 
 		return !fp->error;
-	} else
-		return GL_TRUE;
+	} else {
+		struct r300_vertex_program *vp = r300SelectVertexShader(ctx);
+		if (!vp->translated)
+			r300TranslateVertexShader(vp);
+
+		return !vp->error;
+	}
 }
 
 void r300InitShaderFuncs(struct dd_function_table *functions)
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index e0996383eb..8c228ab5e7 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -2045,6 +2045,7 @@ void r300UpdateShaders(r300ContextPtr rmesa)
 	}
 
 	if (rmesa->radeon.NewGLState && rmesa->options.hw_tcl_enabled) {
+		struct r300_vertex_program *vp;
 		int i;
 		for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
 			rmesa->temp_attrib[i] =
@@ -2060,8 +2061,11 @@ void r300UpdateShaders(r300ContextPtr rmesa)
 			    rmesa->temp_attrib[i];
 		}
 
-		r300SelectVertexShader(rmesa);
-		r300SwitchFallback(ctx, R300_FALLBACK_VERTEX_PROGRAM, rmesa->selected_vp->error);
+		vp = r300SelectVertexShader(ctx);
+		if (!vp->translated)
+			r300TranslateVertexShader(vp);
+
+		r300SwitchFallback(ctx, R300_FALLBACK_VERTEX_PROGRAM, vp->error);
 	}
 
 	fp = r300SelectFragmentShader(ctx);
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index 25c42814ec..326183bda4 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -1027,7 +1027,7 @@ static void t_inputs_outputs(struct r300_vertex_program *vp)
 	}
 }
 
-static void r300TranslateVertexShader(struct r300_vertex_program *vp)
+void r300TranslateVertexShader(struct r300_vertex_program *vp)
 {
 	struct prog_instruction *vpi = vp->Base->Base.Instructions;
 	int i;
@@ -1524,7 +1524,6 @@ static struct r300_vertex_program *build_program(GLcontext *ctx,
 
 	assert(prog->NumInstructions);
 	vp->num_temporaries = prog->NumTemporaries;
-	r300TranslateVertexShader(vp);
 
 	return vp;
 }
@@ -1538,9 +1537,9 @@ static void add_outputs(struct r300_vertex_program_key *key, GLint vert)
 	key->OutputsAdded |= 1 << vert;
 }
 
-void r300SelectVertexShader(r300ContextPtr r300)
+struct r300_vertex_program * r300SelectVertexShader(GLcontext *ctx)
 {
-	GLcontext *ctx = ctx = r300->radeon.glCtx;
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	GLuint InputsRead;
 	struct r300_vertex_program_key wanted_key = { 0 };
 	GLint i;
@@ -1596,14 +1595,14 @@ void r300SelectVertexShader(r300ContextPtr r300)
 	for (vp = vpc->progs; vp; vp = vp->next)
 		if (_mesa_memcmp(&vp->key, &wanted_key, sizeof(wanted_key))
 		    == 0) {
-			r300->selected_vp = vp;
-			return;
+			return r300->selected_vp = vp;
 		}
 
 	vp = build_program(ctx, &wanted_key, &vpc->mesa_program, wpos_idx);
 	vp->next = vpc->progs;
 	vpc->progs = vp;
-	r300->selected_vp = vp;
+
+	return r300->selected_vp = vp;
 }
 
 #define bump_vpu_count(ptr, new_count)   do { \
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.h b/src/mesa/drivers/dri/r300/r300_vertprog.h
index b552e3fb1b..2dab11c337 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.h
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.h
@@ -34,4 +34,8 @@
 
 void r300SetupVertexProgram(r300ContextPtr rmesa);
 
+struct r300_vertex_program * r300SelectVertexShader(GLcontext *ctx);
+
+void r300TranslateVertexShader(struct r300_vertex_program *vp);
+
 #endif
-- 
cgit v1.2.3


From 7360f83364b407e949529eeca9f8c421d2da3ded Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 4 Jul 2009 13:17:35 +0200
Subject: r300: move fragment program selection before vertex program selection

Prepare for wpos and fogc handling rewrite.
---
 src/mesa/drivers/dri/r300/r300_state.c | 44 +++++++++++++++++++---------------
 1 file changed, 25 insertions(+), 19 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index 8c228ab5e7..344f021acf 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -2035,7 +2035,6 @@ static void r300ResetHwState(r300ContextPtr r300)
 void r300UpdateShaders(r300ContextPtr rmesa)
 {
 	GLcontext *ctx = rmesa->radeon.glCtx;
-	struct r300_fragment_program *fp;
 
 	/* should only happenen once, just after context is created */
 	/* TODO: shouldn't we fallback to sw here? */
@@ -2044,21 +2043,34 @@ void r300UpdateShaders(r300ContextPtr rmesa)
 		return;
 	}
 
-	if (rmesa->radeon.NewGLState && rmesa->options.hw_tcl_enabled) {
+	{
+		struct r300_fragment_program *fp;
+
+		fp = r300SelectFragmentShader(ctx);
+		if (!fp->translated)
+			r300TranslateFragmentShader(ctx, fp);
+
+		r300SwitchFallback(ctx, R300_FALLBACK_FRAGMENT_PROGRAM, fp->error);
+	}
+
+	if (rmesa->options.hw_tcl_enabled) {
 		struct r300_vertex_program *vp;
-		int i;
-		for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
-			rmesa->temp_attrib[i] =
-			    TNL_CONTEXT(ctx)->vb.AttribPtr[i];
-			TNL_CONTEXT(ctx)->vb.AttribPtr[i] =
-			    &rmesa->dummy_attrib[i];
-		}
 
-		_tnl_UpdateFixedFunctionProgram(ctx);
+		if (rmesa->radeon.NewGLState) {
+			int i;
+			for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
+				rmesa->temp_attrib[i] =
+				    TNL_CONTEXT(ctx)->vb.AttribPtr[i];
+				TNL_CONTEXT(ctx)->vb.AttribPtr[i] =
+				    &rmesa->dummy_attrib[i];
+			}
+
+			_tnl_UpdateFixedFunctionProgram(ctx);
 
-		for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
-			TNL_CONTEXT(ctx)->vb.AttribPtr[i] =
-			    rmesa->temp_attrib[i];
+			for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
+				TNL_CONTEXT(ctx)->vb.AttribPtr[i] =
+				    rmesa->temp_attrib[i];
+			}
 		}
 
 		vp = r300SelectVertexShader(ctx);
@@ -2068,12 +2080,6 @@ void r300UpdateShaders(r300ContextPtr rmesa)
 		r300SwitchFallback(ctx, R300_FALLBACK_VERTEX_PROGRAM, vp->error);
 	}
 
-	fp = r300SelectFragmentShader(ctx);
-	if (!fp->translated)
-		r300TranslateFragmentShader(ctx, fp);
-
-	r300SwitchFallback(ctx, R300_FALLBACK_FRAGMENT_PROGRAM, fp->error);
-
 	r300UpdateStateParameters(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
 	rmesa->radeon.NewGLState = 0;
 }
-- 
cgit v1.2.3


From d1e4caa6e2b6a1e20feb97ae51703d5b4b18f70b Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 4 Jul 2009 14:57:42 +0200
Subject: r300: recalculate used inputs and outputs after dead code removal

---
 src/mesa/drivers/dri/r300/radeon_nqssadce.c | 28 ++++++++++++++++++++++++++++
 1 file changed, 28 insertions(+)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/radeon_nqssadce.c b/src/mesa/drivers/dri/r300/radeon_nqssadce.c
index 4a2e1cba40..d2591fa1bd 100644
--- a/src/mesa/drivers/dri/r300/radeon_nqssadce.c
+++ b/src/mesa/drivers/dri/r300/radeon_nqssadce.c
@@ -264,6 +264,32 @@ static void process_instruction(struct nqssadce_state* s)
 	}
 }
 
+static void calculateInputsOutputs(struct gl_program *p)
+{
+	struct prog_instruction *inst;
+	int i, tmp;
+	GLuint InputsRead, OutputsWritten;
+
+	inst = p->Instructions;
+	InputsRead = 0;
+	OutputsWritten = 0;
+	while (inst->Opcode != OPCODE_END)
+	{
+		tmp = _mesa_num_inst_src_regs(inst->Opcode);
+		for (i = 0; i < tmp; ++i) {
+			if (inst->SrcReg[i].File == PROGRAM_INPUT)
+				InputsRead |= 1 << inst->SrcReg[i].Index;
+		}
+
+		if (inst->DstReg.File == PROGRAM_OUTPUT)
+			OutputsWritten |= 1 << inst->DstReg.Index;
+
+		++inst;
+	}
+
+	p->InputsRead = InputsRead;
+	p->OutputsWritten = OutputsWritten;
+}
 
 void radeonNqssaDce(GLcontext *ctx, struct gl_program *p, struct radeon_nqssadce_descr* descr)
 {
@@ -280,4 +306,6 @@ void radeonNqssaDce(GLcontext *ctx, struct gl_program *p, struct radeon_nqssadce
 		s.IP--;
 		process_instruction(&s);
 	}
+
+	calculateInputsOutputs(p);
 }
-- 
cgit v1.2.3


From df5fe747fa08f63b949ba0fd6628060831b562ec Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sun, 5 Jul 2009 02:03:32 +0200
Subject: r300: bind vertex program to fragment program

---
 src/mesa/drivers/dri/r300/r300_context.h  |   4 +-
 src/mesa/drivers/dri/r300/r300_draw.c     |   4 +-
 src/mesa/drivers/dri/r300/r300_state.c    |   4 +-
 src/mesa/drivers/dri/r300/r300_vertprog.c | 205 ++++++++++++++----------------
 4 files changed, 103 insertions(+), 114 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 44211a45b3..6f99e49050 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -409,9 +409,7 @@ struct r300_vertex_program {
 	struct r300_vertex_program *next;
 
 	struct r300_vertex_program_key {
-		GLuint InputsRead;
-		GLuint OutputsWritten;
-		GLuint OutputsAdded;
+		GLuint FpReads;
 	} key;
 	
 	struct r300_vertex_shader_hw_code {
diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c
index 4e8b62f186..5420293b91 100644
--- a/src/mesa/drivers/dri/r300/r300_draw.c
+++ b/src/mesa/drivers/dri/r300/r300_draw.c
@@ -332,7 +332,7 @@ static void r300SetVertexFormat(GLcontext *ctx, const struct gl_client_array *ar
 	{
 		int i, tmp;
 
-		tmp = r300->selected_vp->key.InputsRead;
+		tmp = r300->selected_vp->Base->Base.InputsRead;
 		i = 0;
 		vbuf->num_attribs = 0;
 		while (tmp) {
@@ -428,7 +428,7 @@ static GLboolean r300TryDrawPrims(GLcontext *ctx,
 	if (r300->fallback)
 		return GL_FALSE;
 
-	r300SetupVAP(ctx, r300->selected_vp->key.InputsRead, r300->selected_vp->key.OutputsWritten);
+	r300SetupVAP(ctx, r300->selected_vp->Base->Base.InputsRead, r300->selected_vp->Base->Base.OutputsWritten);
 
 	r300UpdateShaderStates(r300);
 
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index 344f021acf..0f3198e792 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -1456,7 +1456,7 @@ static void r300SetupRSUnit(GLcontext * ctx)
 	hw_tcl_on = r300->options.hw_tcl_enabled;
 
 	if (hw_tcl_on)
-		OutputsWritten.vp_outputs = r300->selected_vp->key.OutputsWritten;
+		OutputsWritten.vp_outputs = r300->selected_vp->Base->Base.OutputsWritten;
 	else
 		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->render_inputs_bitset);
 
@@ -1573,7 +1573,7 @@ static void r500SetupRSUnit(GLcontext * ctx)
 	hw_tcl_on = r300->options.hw_tcl_enabled;
 
 	if (hw_tcl_on)
-		OutputsWritten.vp_outputs = r300->selected_vp->key.OutputsWritten;
+		OutputsWritten.vp_outputs = r300->selected_vp->Base->Base.OutputsWritten;
 	else
 		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->render_inputs_bitset);
 
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index 326183bda4..0b04830350 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -961,10 +961,14 @@ static void t_inputs_outputs(struct r300_vertex_program *vp)
 {
 	int i;
 	int cur_reg;
+	GLuint OutputsWritten, InputsRead;
+
+	OutputsWritten = vp->Base->Base.OutputsWritten;
+	InputsRead = vp->Base->Base.InputsRead;
 
 	cur_reg = -1;
 	for (i = 0; i < VERT_ATTRIB_MAX; i++) {
-		if (vp->key.InputsRead & (1 << i))
+		if (InputsRead & (1 << i))
 			vp->inputs[i] = ++cur_reg;
 		else
 			vp->inputs[i] = -1;
@@ -974,13 +978,13 @@ static void t_inputs_outputs(struct r300_vertex_program *vp)
 	for (i = 0; i < VERT_RESULT_MAX; i++)
 		vp->outputs[i] = -1;
 
-	assert(vp->key.OutputsWritten & (1 << VERT_RESULT_HPOS));
+	assert(OutputsWritten & (1 << VERT_RESULT_HPOS));
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_HPOS)) {
+	if (OutputsWritten & (1 << VERT_RESULT_HPOS)) {
 		vp->outputs[VERT_RESULT_HPOS] = cur_reg++;
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_PSIZ)) {
+	if (OutputsWritten & (1 << VERT_RESULT_PSIZ)) {
 		vp->outputs[VERT_RESULT_PSIZ] = cur_reg++;
 	}
 
@@ -990,39 +994,39 @@ static void t_inputs_outputs(struct r300_vertex_program *vp)
 	 * pretend it does by skipping output index reg so the colors
 	 * get written into appropriate output vectors.
 	 */
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL0)) {
+	if (OutputsWritten & (1 << VERT_RESULT_COL0)) {
 		vp->outputs[VERT_RESULT_COL0] = cur_reg++;
-	} else if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC0) ||
-		vp->key.OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+	} else if (OutputsWritten & (1 << VERT_RESULT_BFC0) ||
+		OutputsWritten & (1 << VERT_RESULT_BFC1)) {
 		cur_reg++;
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL1)) {
+	if (OutputsWritten & (1 << VERT_RESULT_COL1)) {
 		vp->outputs[VERT_RESULT_COL1] = cur_reg++;
-	} else if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC0) ||
-		vp->key.OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+	} else if (OutputsWritten & (1 << VERT_RESULT_BFC0) ||
+		OutputsWritten & (1 << VERT_RESULT_BFC1)) {
 		cur_reg++;
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC0)) {
+	if (OutputsWritten & (1 << VERT_RESULT_BFC0)) {
 		vp->outputs[VERT_RESULT_BFC0] = cur_reg++;
-	} else if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+	} else if (OutputsWritten & (1 << VERT_RESULT_BFC1)) {
 		cur_reg++;
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+	if (OutputsWritten & (1 << VERT_RESULT_BFC1)) {
 		vp->outputs[VERT_RESULT_BFC1] = cur_reg++;
-	} else if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC0)) {
+	} else if (OutputsWritten & (1 << VERT_RESULT_BFC0)) {
 		cur_reg++;
 	}
 
 	for (i = VERT_RESULT_TEX0; i <= VERT_RESULT_TEX7; i++) {
-		if (vp->key.OutputsWritten & (1 << i)) {
+		if (OutputsWritten & (1 << i)) {
 			vp->outputs[i] = cur_reg++;
 		}
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_FOGC)) {
+	if (OutputsWritten & (1 << VERT_RESULT_FOGC)) {
 		vp->outputs[VERT_RESULT_FOGC] = cur_reg++;
 	}
 }
@@ -1255,6 +1259,8 @@ static void insert_wpos(struct r300_vertex_program *vp, struct gl_program *prog,
 	++vpi;
 
 	vpi->Opcode = OPCODE_END;
+
+	prog->OutputsWritten |= 1 << (VERT_RESULT_TEX0 + vp->wpos_idx);
 }
 
 static void pos_as_texcoord(struct r300_vertex_program *vp,
@@ -1448,6 +1454,63 @@ static void translateInsts(struct gl_program *prog)
 	}
 }
 
+#define ADD_OUTPUT(fp_attr, vp_result) \
+	do { \
+		if ((FpReads & (1 << (fp_attr))) && !(prog->OutputsWritten & (1 << (vp_result)))) { \
+			OutputsAdded |= 1 << (vp_result); \
+			count++; \
+		} \
+	} while (0)
+
+static void addArtificialOutputs(GLcontext *ctx, struct gl_program *prog)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	GLuint OutputsAdded, FpReads;
+	int i, count;
+
+	OutputsAdded = count = 0;
+	FpReads = r300->selected_fp->Base->InputsRead;
+
+	ADD_OUTPUT(FRAG_ATTRIB_COL0, VERT_RESULT_COL0);
+	ADD_OUTPUT(FRAG_ATTRIB_COL1, VERT_RESULT_COL1);
+
+	for (i = 0; i < 7; ++i) {
+		ADD_OUTPUT(FRAG_ATTRIB_TEX0 + i, VERT_RESULT_TEX0 + i);
+	}
+
+	/* Some outputs may be artificially added, to match the inputs of the fragment program.
+	 * Issue 16 of vertex program spec says that all vertex attributes that are unwritten by
+	 * vertex program are undefined, so just use MOV [vertex_result], CONST[0]
+	 */
+	if (count > 0) {
+		struct prog_instruction *inst;
+
+		_mesa_insert_instructions(prog, prog->NumInstructions - 1, count);
+		inst = &prog->Instructions[prog->NumInstructions - 1 - count];
+
+		for (i = 0; i < VERT_RESULT_MAX; ++i) {
+			if (OutputsAdded & (1 << i)) {
+				inst->Opcode = OPCODE_MOV;
+
+				inst->DstReg.File = PROGRAM_OUTPUT;
+				inst->DstReg.Index = i;
+				inst->DstReg.WriteMask = WRITEMASK_XYZW;
+				inst->DstReg.CondMask = COND_TR;
+
+				inst->SrcReg[0].File = PROGRAM_CONSTANT;
+				inst->SrcReg[0].Index = 0;
+				inst->SrcReg[0].Swizzle = SWIZZLE_XYZW;
+
+				++inst;
+			}
+		}
+
+		prog->OutputsWritten |= OutputsAdded;
+	}
+}
+
+#undef ADD_OUTPUT
+
 static struct r300_vertex_program *build_program(GLcontext *ctx,
 						 struct r300_vertex_program_key *wanted_key,
 						 const struct gl_vertex_program *mesa_vp,
@@ -1477,42 +1540,7 @@ static struct r300_vertex_program *build_program(GLcontext *ctx,
 		pos_as_texcoord(vp, prog);
 	}
 
-	/* Some outputs may be artificially added, to match the inputs of the fragment program.
-	 * Issue 16 of vertex program spec says that all vertex attributes that are unwritten by
-	 * vertex program are undefined, so just use MOV [vertex_result], CONST[0]
-	 */
-	{
-		int i, count = 0;
-		for (i = 0; i < VERT_RESULT_MAX; ++i) {
-			if (vp->key.OutputsAdded & (1 << i)) {
-				++count;
-			}
-		}
-
-		if (count > 0) {
-			struct prog_instruction *inst;
-
-			_mesa_insert_instructions(prog, prog->NumInstructions - 1, count);
-			inst = &prog->Instructions[prog->NumInstructions - 1 - count];
-
-			for (i = 0; i < VERT_RESULT_MAX; ++i) {
-				if (vp->key.OutputsAdded & (1 << i)) {
-					inst->Opcode = OPCODE_MOV;
-
-					inst->DstReg.File = PROGRAM_OUTPUT;
-					inst->DstReg.Index = i;
-					inst->DstReg.WriteMask = WRITEMASK_XYZW;
-					inst->DstReg.CondMask = COND_TR;
-
-					inst->SrcReg[0].File = PROGRAM_CONSTANT;
-					inst->SrcReg[0].Index = 0;
-					inst->SrcReg[0].Swizzle = SWIZZLE_XYZW;
-
-					++inst;
-				}
-			}
-		}
-	}
+	addArtificialOutputs(ctx, prog);
 
 	translateInsts(prog);
 
@@ -1528,76 +1556,39 @@ static struct r300_vertex_program *build_program(GLcontext *ctx,
 	return vp;
 }
 
-static void add_outputs(struct r300_vertex_program_key *key, GLint vert)
-{
-	if (key->OutputsWritten & (1 << vert))
-		return;
-
-	key->OutputsWritten |= 1 << vert;
-	key->OutputsAdded |= 1 << vert;
-}
-
 struct r300_vertex_program * r300SelectVertexShader(GLcontext *ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-	GLuint InputsRead;
 	struct r300_vertex_program_key wanted_key = { 0 };
-	GLint i;
 	struct r300_vertex_program_cont *vpc;
 	struct r300_vertex_program *vp;
 	GLint wpos_idx;
 
 	vpc = (struct r300_vertex_program_cont *)ctx->VertexProgram._Current;
-	wanted_key.InputsRead = vpc->mesa_program.Base.InputsRead;
-	wanted_key.OutputsWritten = vpc->mesa_program.Base.OutputsWritten;
-	InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
-
-	wpos_idx = -1;
-	if (InputsRead & FRAG_BIT_WPOS) {
-		for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
-			if (!(InputsRead & (FRAG_BIT_TEX0 << i)))
-				break;
+	wanted_key.FpReads = r300->selected_fp->Base->InputsRead;
 
-		if (i == ctx->Const.MaxTextureUnits) {
-			fprintf(stderr, "\tno free texcoord found\n");
-			_mesa_exit(-1);
+	for (vp = vpc->progs; vp; vp = vp->next)
+		if (_mesa_memcmp(&vp->key, &wanted_key, sizeof(wanted_key))
+		    == 0) {
+			return r300->selected_vp = vp;
 		}
 
-		wanted_key.OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i);
-		wpos_idx = i;
-	}
-
-	if (vpc->mesa_program.IsPositionInvariant) {
-		wanted_key.InputsRead |= (1 << VERT_ATTRIB_POS);
-		wanted_key.OutputsWritten |= (1 << VERT_RESULT_HPOS);
-	} else {
-		add_outputs(&wanted_key, VERT_RESULT_HPOS);
-	}
-
-	if (InputsRead & FRAG_BIT_COL0) {
-		add_outputs(&wanted_key, VERT_RESULT_COL0);
-	}
+	wpos_idx = -1;
+	if (wanted_key.FpReads & FRAG_BIT_WPOS) {
+		GLint i;
 
-	if (InputsRead & FRAG_BIT_COL1) {
-		add_outputs(&wanted_key, VERT_RESULT_COL1);
-	}
+		for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
+			if (!(wanted_key.FpReads & (FRAG_BIT_TEX(i))))
+				break;
 
-	if (InputsRead & FRAG_BIT_FOGC) {
-		add_outputs(&wanted_key, VERT_RESULT_FOGC);
-	}
+			if (i == ctx->Const.MaxTextureUnits) {
+				fprintf(stderr, "\tno free texcoord found\n");
+				_mesa_exit(-1);
+			}
 
-	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-		if (InputsRead & (FRAG_BIT_TEX0 << i)) {
-			add_outputs(&wanted_key, VERT_RESULT_TEX0 + i);
-		}
+			wpos_idx = i;
 	}
 
-	for (vp = vpc->progs; vp; vp = vp->next)
-		if (_mesa_memcmp(&vp->key, &wanted_key, sizeof(wanted_key))
-		    == 0) {
-			return r300->selected_vp = vp;
-		}
-
 	vp = build_program(ctx, &wanted_key, &vpc->mesa_program, wpos_idx);
 	vp->next = vpc->progs;
 	vpc->progs = vp;
@@ -1663,8 +1654,8 @@ void r300SetupVertexProgram(r300ContextPtr rmesa)
 	r300EmitVertexProgram(rmesa, R300_PVS_CODE_START, &(prog->hw_code));
 	inst_count = (prog->hw_code.length / 4) - 1;
 
-	r300VapCntl(rmesa, _mesa_bitcount(prog->key.InputsRead),
-				 _mesa_bitcount(prog->key.OutputsWritten), prog->num_temporaries);
+	r300VapCntl(rmesa, _mesa_bitcount(prog->Base->Base.InputsRead),
+				 _mesa_bitcount(prog->Base->Base.OutputsWritten), prog->num_temporaries);
 
 	R300_STATECHANGE(rmesa, pvs);
 	rmesa->hw.pvs.cmd[R300_PVS_CNTL_1] = (0 << R300_PVS_FIRST_INST_SHIFT) | (inst_count << R300_PVS_XYZW_VALID_INST_SHIFT) |
-- 
cgit v1.2.3


From f79ef95df4f19124c24e59583bf9fb1e347d8f2b Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sun, 5 Jul 2009 02:32:51 +0200
Subject: r300: rewrite FOGC and HPOS attribs handling

Rewrite vertex and fragment programs so that we don't have to do any hacks on lower level.
---
 src/mesa/drivers/dri/r300/r300_context.h         |  8 ++-
 src/mesa/drivers/dri/r300/r300_emit.c            |  9 ---
 src/mesa/drivers/dri/r300/r300_fragprog_common.c | 62 ++++++++++++++++++-
 src/mesa/drivers/dri/r300/r300_shader.c          |  3 +-
 src/mesa/drivers/dri/r300/r300_state.c           | 53 ----------------
 src/mesa/drivers/dri/r300/r300_swtcl.c           | 38 +++++-------
 src/mesa/drivers/dri/r300/r300_vertprog.c        | 79 ++++++++++++++----------
 7 files changed, 130 insertions(+), 122 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 6f99e49050..32104b8158 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -410,6 +410,8 @@ struct r300_vertex_program {
 
 	struct r300_vertex_program_key {
 		GLuint FpReads;
+		GLuint FogAttr;
+		GLuint WPosAttr;
 	} key;
 	
 	struct r300_vertex_shader_hw_code {
@@ -425,7 +427,6 @@ struct r300_vertex_program {
 
 	int pos_end;
 	int num_temporaries;	/* Number of temp vars used by program */
-	int wpos_idx;
 	int inputs[VERT_ATTRIB_MAX];
 	int outputs[VERT_RESULT_MAX];
 };
@@ -560,6 +561,11 @@ struct r300_fragment_program {
 	GLuint optimization;
 
 	struct r300_fragment_program *next;
+
+	/* attribute that we are sending the WPOS in */
+	gl_frag_attrib wpos_attr;
+	/* attribute that we are sending the fog coordinate in */
+	gl_frag_attrib fog_attr;
 };
 
 struct r300_fragment_program_cont {
diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c
index c3817721dc..707d1284ed 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.c
+++ b/src/mesa/drivers/dri/r300/r300_emit.c
@@ -116,15 +116,6 @@ GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint vp_writes, GLuint fp_reads)
 		}
 	}
 
-	if (fp_reads & FRAG_BIT_WPOS) {
-		ret |= (4 << (3 * first_free_texcoord));
-		++first_free_texcoord;
-	}
-
-	if (vp_writes & (1 << VERT_RESULT_FOGC) && fp_reads & FRAG_BIT_FOGC) {
-		ret |= 4 << (3 * first_free_texcoord);
-	}
-
 	if (first_free_texcoord > 8) {
 		fprintf(stderr, "\tout of free texcoords\n");
 		_mesa_exit(-1);
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.c b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
index b25cf24007..e90be9b7f8 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_common.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
@@ -69,8 +69,10 @@ static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler)
 {
 	GLuint InputsRead = compiler->fp->Base->InputsRead;
 
-	if (!(InputsRead & FRAG_BIT_WPOS))
+	if (!(InputsRead & FRAG_BIT_WPOS)) {
+		compiler->fp->wpos_attr = FRAG_ATTRIB_MAX;
 		return;
+	}
 
 	static gl_state_index tokens[STATE_LENGTH] = {
 		STATE_INTERNAL, STATE_R300_WINDOW_DIMENSION, 0, 0, 0
@@ -78,10 +80,23 @@ static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler)
 	struct prog_instruction *fpi;
 	GLuint window_index;
 	int i = 0;
+
+	for (i = FRAG_ATTRIB_TEX0; i <= FRAG_ATTRIB_TEX7; ++i)
+	{
+		if (!(InputsRead & (1 << i))) {
+			InputsRead &= ~(1 << FRAG_ATTRIB_WPOS);
+			InputsRead |= 1 << i;
+			compiler->fp->Base->InputsRead = InputsRead;
+			compiler->fp->wpos_attr = i;
+			break;
+		}
+	}
+
 	GLuint tempregi = _mesa_find_free_register(compiler->program, PROGRAM_TEMPORARY);
 
 	_mesa_insert_instructions(compiler->program, 0, 3);
 	fpi = compiler->program->Instructions;
+	i = 0;
 
 	/* perspective divide */
 	fpi[i].Opcode = OPCODE_RCP;
@@ -92,7 +107,7 @@ static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler)
 	fpi[i].DstReg.CondMask = COND_TR;
 
 	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
-	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
+	fpi[i].SrcReg[0].Index = compiler->fp->wpos_attr;
 	fpi[i].SrcReg[0].Swizzle = SWIZZLE_WWWW;
 	i++;
 
@@ -104,7 +119,7 @@ static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler)
 	fpi[i].DstReg.CondMask = COND_TR;
 
 	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
-	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
+	fpi[i].SrcReg[0].Index = compiler->fp->wpos_attr;
 	fpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
 
 	fpi[i].SrcReg[1].File = PROGRAM_TEMPORARY;
@@ -147,6 +162,45 @@ static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler)
 	}
 }
 
+static void rewriteFog(struct r300_fragment_program_compiler *compiler)
+{
+	struct r300_fragment_program *fp = compiler->fp;
+	GLuint InputsRead;
+	int i;
+
+	InputsRead = fp->Base->InputsRead;
+
+	if (!(InputsRead & FRAG_BIT_FOGC)) {
+		fp->fog_attr = FRAG_ATTRIB_MAX;
+		return;
+	}
+
+	for (i = FRAG_ATTRIB_TEX0; i <= FRAG_ATTRIB_TEX7; ++i)
+	{
+		if (!(InputsRead & (1 << i))) {
+			InputsRead &= ~(1 << FRAG_ATTRIB_FOGC);
+			InputsRead |= 1 << i;
+			fp->Base->InputsRead = InputsRead;
+			fp->fog_attr = i;
+			break;
+		}
+	}
+
+	{
+		struct prog_instruction *inst;
+
+		inst = compiler->program->Instructions;
+		while (inst->Opcode != OPCODE_END) {
+			const int src_regs = _mesa_num_inst_src_regs(inst->Opcode);
+			for (i = 0; i < src_regs; ++i) {
+				if (inst->SrcReg[i].File == PROGRAM_INPUT && inst->SrcReg[i].Index == FRAG_ATTRIB_FOGC)
+					inst->SrcReg[i].Index = fp->fog_attr;
+			}
+			++inst;
+		}
+	}
+}
+
 static GLuint build_dtm(GLuint depthmode)
 {
 	switch(depthmode) {
@@ -204,6 +258,8 @@ void r300TranslateFragmentShader(GLcontext *ctx, struct r300_fragment_program *f
 
 	insert_WPOS_trailer(&compiler);
 
+	rewriteFog(&compiler);
+
 	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
 		struct radeon_program_transformation transformations[] = {
 			{ &r500_transform_TEX, &compiler },
diff --git a/src/mesa/drivers/dri/r300/r300_shader.c b/src/mesa/drivers/dri/r300/r300_shader.c
index 854eb5d80a..62228a3786 100644
--- a/src/mesa/drivers/dri/r300/r300_shader.c
+++ b/src/mesa/drivers/dri/r300/r300_shader.c
@@ -66,8 +66,7 @@ static struct gl_program *r300NewProgram(GLcontext * ctx, GLenum target,
 	case GL_VERTEX_STATE_PROGRAM_NV:
 	case GL_VERTEX_PROGRAM_ARB:
 		vp = CALLOC_STRUCT(r300_vertex_program_cont);
-		return _mesa_init_vertex_program(ctx, &vp->mesa_program,
-						 target, id);
+		return _mesa_init_vertex_program(ctx, &vp->mesa_program, target, id);
 
 	case GL_FRAGMENT_PROGRAM_NV:
 	case GL_FRAGMENT_PROGRAM_ARB:
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index 0f3198e792..bae2f0f3cb 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -1519,29 +1519,6 @@ static void r300SetupRSUnit(GLcontext * ctx)
 		++fp_reg;
 	}
 
-	if (InputsRead & FRAG_BIT_WPOS) {
-		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3) | R300_RS_TEX_PTR(rs_tex_count);
-		r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R300_RS_INST_TEX_ID(tex_ip) | R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_reg);
-		InputsRead &= ~FRAG_BIT_WPOS;
-		rs_tex_count += 4;
-		++tex_ip;
-		++fp_reg;
-	}
-
-	if (InputsRead & FRAG_BIT_FOGC) {
-		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_FOGC, _TNL_ATTRIB_FOG)) {
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= R300_RS_SEL_S(0) | R300_RS_SEL_T(R300_RS_SEL_K0) | R300_RS_SEL_R(R300_RS_SEL_K0);
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= R300_RS_SEL_Q(R300_RS_SEL_K1) | R300_RS_TEX_PTR(rs_tex_count);
-			r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R300_RS_INST_TEX_ID(tex_ip) | R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_reg);
-			InputsRead &= ~FRAG_BIT_FOGC;
-			rs_tex_count += 4;
-			++tex_ip;
-			++fp_reg;
-		} else {
-			WARN_ONCE("fragprog wants fogc, vp doesn't provide it\n");
-		}
-	}
-
 	/* Setup default color if no color or tex was set */
 	if (rs_tex_count == 0 && col_ip == 0) {
 		r300->hw.rr.cmd[R300_RR_INST_0] = R300_RS_INST_COL_ID(0) | R300_RS_INST_COL_ADDR(0);
@@ -1640,36 +1617,6 @@ static void r500SetupRSUnit(GLcontext * ctx)
 		++fp_reg;
 	}
 
-	if (InputsRead & FRAG_BIT_WPOS) {
-		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= ((rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT) |
-			((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) |
-			((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) |
-			((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT);
-
-		r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg);
-		InputsRead &= ~FRAG_BIT_WPOS;
-		rs_tex_count += 4;
-		++tex_ip;
-		++fp_reg;
-	}
-
-	if (InputsRead & FRAG_BIT_FOGC) {
-		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_FOGC, _TNL_ATTRIB_FOG)) {
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= (rs_tex_count << R500_RS_IP_TEX_PTR_S_SHIFT) |
-				(R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) |
-				(R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) |
-				(R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT);
-
-			r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg);
-			InputsRead &= ~FRAG_BIT_FOGC;
-			rs_tex_count += 4;
-			++tex_ip;
-			++fp_reg;
-		} else {
-			WARN_ONCE("fragprog wants fogc, vp doesn't provide it\n");
-		}
-	}
-
 	/* Setup default color if no color or tex was set */
 	if (rs_tex_count == 0 && col_ip == 0) {
 		r300->hw.rr.cmd[R300_RR_INST_0] = R500_RS_INST_COL_ID(0) | R500_RS_INST_COL_ADDR(0);
diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c
index db4ccce6f1..d73508d36e 100644
--- a/src/mesa/drivers/dri/r300/r300_swtcl.c
+++ b/src/mesa/drivers/dri/r300/r300_swtcl.c
@@ -192,31 +192,23 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead,  GLuint *_
 		}
 	}
 
-	/* RS can't put fragment position on the pixel stack, so stuff it in texcoord if needed */
-	if (fp_reads & FRAG_BIT_WPOS) {
-		if (first_free_tex >= ctx->Const.MaxTextureUnits) {
-			fprintf(stderr, "\tout of free texcoords to write w pos\n");
-			_mesa_exit(-1);
-		}
+	if (rmesa->selected_fp->wpos_attr != FRAG_ATTRIB_MAX) {
+		int tex_id = rmesa->selected_fp->wpos_attr - FRAG_ATTRIB_TEX0;
 
-		InputsRead |= 1 << (VERT_ATTRIB_TEX0 + first_free_tex);
-		OutputsWritten |= 1 << (VERT_RESULT_TEX0 + first_free_tex);
-		EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F );
-		ADD_ATTR(VERT_ATTRIB_POS, R300_DATA_TYPE_FLOAT_4, SWTCL_OVM_TEX(first_free_tex), SWIZZLE_XYZW, MASK_XYZW, 0);
-		++first_free_tex;
+		VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_POS];
+		RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id);
 	}
 
-	if (fp_reads & FRAG_BIT_FOGC) {
-		if (first_free_tex >= ctx->Const.MaxTextureUnits) {
-			fprintf(stderr, "\tout of free texcoords to write fog coordinate\n");
-			_mesa_exit(-1);
-		}
+	if (rmesa->selected_fp->fog_attr != FRAG_ATTRIB_MAX) {
+		int tex_id = rmesa->selected_fp->fog_attr - FRAG_ATTRIB_TEX0;
 
-		InputsRead |= 1 << VERT_ATTRIB_FOG;
-		OutputsWritten |= 1 << VERT_RESULT_FOGC;
-		GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO);
-		EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1F );
-		ADD_ATTR(VERT_ATTRIB_FOG, R300_DATA_TYPE_FLOAT_1, SWTCL_OVM_TEX(first_free_tex), swiz, MASK_XYZW, 0);
+		VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG];
+		RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id);
+	}
+
+	if (first_free_tex >= ctx->Const.MaxTextureUnits) {
+		fprintf(stderr, "\tout of free texcoords to write fog coordinate\n");
+		_mesa_exit(-1);
 	}
 
 	R300_NEWPRIM(rmesa);
@@ -497,11 +489,13 @@ void r300RenderStart(GLcontext *ctx)
 	r300ContextPtr rmesa = R300_CONTEXT( ctx );
 
 	r300ChooseRenderState(ctx);
+
+	r300UpdateShaders(rmesa);
+
 	r300PrepareVertices(ctx);
 
 	r300ValidateBuffers(ctx);
 
-	r300UpdateShaders(rmesa);
 	r300UpdateShaderStates(rmesa);
 
 	r300EmitCacheFlush(rmesa);
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index 0b04830350..6f9b4db40f 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -1223,8 +1223,7 @@ void r300TranslateVertexShader(struct r300_vertex_program *vp)
 	}
 }
 
-static void insert_wpos(struct r300_vertex_program *vp, struct gl_program *prog,
-			GLuint temp_index)
+static void insert_wpos(struct gl_program *prog, GLuint temp_index, int tex_id)
 {
 	struct prog_instruction *vpi;
 
@@ -1248,7 +1247,7 @@ static void insert_wpos(struct r300_vertex_program *vp, struct gl_program *prog,
 	vpi->Opcode = OPCODE_MOV;
 
 	vpi->DstReg.File = PROGRAM_OUTPUT;
-	vpi->DstReg.Index = VERT_RESULT_TEX0 + vp->wpos_idx;
+	vpi->DstReg.Index = VERT_RESULT_TEX0 + tex_id;
 	vpi->DstReg.WriteMask = WRITEMASK_XYZW;
 	vpi->DstReg.CondMask = COND_TR;
 
@@ -1259,12 +1258,9 @@ static void insert_wpos(struct r300_vertex_program *vp, struct gl_program *prog,
 	++vpi;
 
 	vpi->Opcode = OPCODE_END;
-
-	prog->OutputsWritten |= 1 << (VERT_RESULT_TEX0 + vp->wpos_idx);
 }
 
-static void pos_as_texcoord(struct r300_vertex_program *vp,
-			    struct gl_program *prog)
+static void pos_as_texcoord(struct gl_program *prog, int tex_id)
 {
 	struct prog_instruction *vpi;
 	GLuint tempregi = prog->NumTemporaries;
@@ -1278,7 +1274,37 @@ static void pos_as_texcoord(struct r300_vertex_program *vp,
 		}
 	}
 
-	insert_wpos(vp, prog, tempregi);
+	insert_wpos(prog, tempregi, tex_id);
+
+	prog->OutputsWritten |= 1 << (VERT_RESULT_TEX0 + tex_id);
+}
+
+/**
+ @TODO
+  We can put X001 swizzle only if input components are directly mapped from output components.
+  For some insts we need to skip source swizzles and add: MOV OUTPUT[fog_attr].yzw, CONST[0].0001
+ */
+static void fog_as_texcoord(struct gl_program *prog, int tex_id)
+{
+	struct prog_instruction *vpi;
+	int i;
+
+	vpi = prog->Instructions;
+	while (vpi->Opcode != OPCODE_END) {
+		if (vpi->DstReg.File == PROGRAM_OUTPUT && vpi->DstReg.Index == VERT_RESULT_FOGC) {
+			vpi->DstReg.Index = VERT_RESULT_TEX0 + tex_id;
+			vpi->DstReg.WriteMask = WRITEMASK_XYZW;
+
+			for (i = 0; i < _mesa_num_inst_src_regs(vpi->Opcode); ++i) {
+				vpi->SrcReg[i].Swizzle = combineSwizzles(vpi->SrcReg[i].Swizzle, SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE);
+			}
+		}
+
+		++vpi;
+	}
+
+	prog->OutputsWritten &= ~(1 << VERT_RESULT_FOGC);
+	prog->OutputsWritten |= 1 << (VERT_RESULT_TEX0 + tex_id);
 }
 
 static int translateABS(struct gl_program *prog, int pos)
@@ -1513,16 +1539,15 @@ static void addArtificialOutputs(GLcontext *ctx, struct gl_program *prog)
 
 static struct r300_vertex_program *build_program(GLcontext *ctx,
 						 struct r300_vertex_program_key *wanted_key,
-						 const struct gl_vertex_program *mesa_vp,
-						 GLint wpos_idx)
+						 const struct gl_vertex_program *mesa_vp)
 {
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	struct r300_vertex_program *vp;
 	struct gl_program *prog;
 
 	vp = _mesa_calloc(sizeof(*vp));
 	vp->Base = (struct gl_vertex_program *) _mesa_clone_program(ctx, &mesa_vp->Base);
 	_mesa_memcpy(&vp->key, wanted_key, sizeof(vp->key));
-	vp->wpos_idx = wpos_idx;
 
 	prog = &vp->Base->Base;
 
@@ -1532,12 +1557,16 @@ static struct r300_vertex_program *build_program(GLcontext *ctx,
 		fflush(stdout);
 	}
 
-	if (mesa_vp->IsPositionInvariant) {
+	if (vp->Base->IsPositionInvariant) {
 		_mesa_insert_mvp_code(ctx, vp->Base);
 	}
 
-	if (wpos_idx > -1) {
-		pos_as_texcoord(vp, prog);
+	if (r300->selected_fp->wpos_attr != FRAG_ATTRIB_MAX) {
+		pos_as_texcoord(&vp->Base->Base, r300->selected_fp->wpos_attr - FRAG_ATTRIB_TEX0);
+	}
+
+	if (r300->selected_fp->fog_attr != FRAG_ATTRIB_MAX) {
+		fog_as_texcoord(&vp->Base->Base, r300->selected_fp->fog_attr - FRAG_ATTRIB_TEX0);
 	}
 
 	addArtificialOutputs(ctx, prog);
@@ -1562,34 +1591,20 @@ struct r300_vertex_program * r300SelectVertexShader(GLcontext *ctx)
 	struct r300_vertex_program_key wanted_key = { 0 };
 	struct r300_vertex_program_cont *vpc;
 	struct r300_vertex_program *vp;
-	GLint wpos_idx;
 
 	vpc = (struct r300_vertex_program_cont *)ctx->VertexProgram._Current;
 	wanted_key.FpReads = r300->selected_fp->Base->InputsRead;
+	wanted_key.FogAttr = r300->selected_fp->fog_attr;
+	wanted_key.WPosAttr = r300->selected_fp->wpos_attr;
 
-	for (vp = vpc->progs; vp; vp = vp->next)
+	for (vp = vpc->progs; vp; vp = vp->next) {
 		if (_mesa_memcmp(&vp->key, &wanted_key, sizeof(wanted_key))
 		    == 0) {
 			return r300->selected_vp = vp;
 		}
-
-	wpos_idx = -1;
-	if (wanted_key.FpReads & FRAG_BIT_WPOS) {
-		GLint i;
-
-		for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
-			if (!(wanted_key.FpReads & (FRAG_BIT_TEX(i))))
-				break;
-
-			if (i == ctx->Const.MaxTextureUnits) {
-				fprintf(stderr, "\tno free texcoord found\n");
-				_mesa_exit(-1);
-			}
-
-			wpos_idx = i;
 	}
 
-	vp = build_program(ctx, &wanted_key, &vpc->mesa_program, wpos_idx);
+	vp = build_program(ctx, &wanted_key, &vpc->mesa_program);
 	vp->next = vpc->progs;
 	vpc->progs = vp;
 
-- 
cgit v1.2.3


From e43cc28c1b6face903f3c977d6eb887335bec886 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 4 Jul 2009 16:50:25 +0200
Subject: r300: move depth output rewrite out of NQSSADCE

---
 src/mesa/drivers/dri/r300/r300_fragprog_common.c | 48 ++++++++++++++++++++++--
 src/mesa/drivers/dri/r300/radeon_nqssadce.c      | 41 +-------------------
 src/mesa/drivers/dri/r300/radeon_nqssadce.h      |  6 +--
 3 files changed, 46 insertions(+), 49 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.c b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
index e90be9b7f8..d5ef18b2ad 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_common.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
@@ -239,6 +239,46 @@ static void build_state(
 	}
 }
 
+static void rewrite_depth_out(struct gl_program *prog)
+{
+	struct prog_instruction *inst;
+
+	for (inst = prog->Instructions; inst->Opcode != OPCODE_END; ++inst) {
+		if (inst->DstReg.File != PROGRAM_OUTPUT || inst->DstReg.Index != FRAG_RESULT_DEPTH)
+			continue;
+
+		if (inst->DstReg.WriteMask & WRITEMASK_Z) {
+			inst->DstReg.WriteMask = WRITEMASK_W;
+		} else {
+			inst->DstReg.WriteMask = 0;
+			continue;
+		}
+
+		switch (inst->Opcode) {
+			case OPCODE_FRC:
+			case OPCODE_MOV:
+				inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
+				break;
+			case OPCODE_ADD:
+			case OPCODE_MAX:
+			case OPCODE_MIN:
+			case OPCODE_MUL:
+				inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
+				inst->SrcReg[1] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[1]);
+				break;
+			case OPCODE_CMP:
+			case OPCODE_MAD:
+				inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
+				inst->SrcReg[1] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[1]);
+				inst->SrcReg[2] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[2]);
+				break;
+			default:
+				// Scalar instructions needn't be reswizzled
+				break;
+		}
+	}
+}
+
 void r300TranslateFragmentShader(GLcontext *ctx, struct r300_fragment_program *fp)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
@@ -260,6 +300,8 @@ void r300TranslateFragmentShader(GLcontext *ctx, struct r300_fragment_program *f
 
 	rewriteFog(&compiler);
 
+	rewrite_depth_out(compiler.program);
+
 	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
 		struct radeon_program_transformation transformations[] = {
 			{ &r500_transform_TEX, &compiler },
@@ -287,16 +329,14 @@ void r300TranslateFragmentShader(GLcontext *ctx, struct r300_fragment_program *f
 		struct radeon_nqssadce_descr nqssadce = {
 			.Init = &nqssadce_init,
 			.IsNativeSwizzle = &r500FPIsNativeSwizzle,
-			.BuildSwizzle = &r500FPBuildSwizzle,
-			.RewriteDepthOut = GL_TRUE
+			.BuildSwizzle = &r500FPBuildSwizzle
 		};
 		radeonNqssaDce(ctx, compiler.program, &nqssadce);
 	} else {
 		struct radeon_nqssadce_descr nqssadce = {
 			.Init = &nqssadce_init,
 			.IsNativeSwizzle = &r300FPIsNativeSwizzle,
-			.BuildSwizzle = &r300FPBuildSwizzle,
-			.RewriteDepthOut = GL_TRUE
+			.BuildSwizzle = &r300FPBuildSwizzle
 		};
 		radeonNqssaDce(ctx, compiler.program, &nqssadce);
 	}
diff --git a/src/mesa/drivers/dri/r300/radeon_nqssadce.c b/src/mesa/drivers/dri/r300/radeon_nqssadce.c
index d2591fa1bd..82dfc31dd0 100644
--- a/src/mesa/drivers/dri/r300/radeon_nqssadce.c
+++ b/src/mesa/drivers/dri/r300/radeon_nqssadce.c
@@ -56,7 +56,7 @@ static struct register_state *get_reg_state(struct nqssadce_state* s, GLuint fil
  *
  * @note Works correctly only for X, Y, Z, W swizzles, not for constant swizzles.
  */
-static struct prog_src_register lmul_swizzle(GLuint swizzle, struct prog_src_register srcreg)
+struct prog_src_register lmul_swizzle(GLuint swizzle, struct prog_src_register srcreg)
 {
 	struct prog_src_register tmp = srcreg;
 	int i;
@@ -121,40 +121,6 @@ static struct prog_instruction* track_used_srcreg(struct nqssadce_state* s,
 	return inst;
 }
 
-
-static void rewrite_depth_out(struct prog_instruction *inst)
-{
-	if (inst->DstReg.WriteMask & WRITEMASK_Z) {
-		inst->DstReg.WriteMask = WRITEMASK_W;
-	} else {
-		inst->DstReg.WriteMask = 0;
-		return;
-	}
-
-	switch (inst->Opcode) {
-	case OPCODE_FRC:
-	case OPCODE_MOV:
-		inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
-		break;
-	case OPCODE_ADD:
-	case OPCODE_MAX:
-	case OPCODE_MIN:
-	case OPCODE_MUL:
-		inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
-		inst->SrcReg[1] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[1]);
-		break;
-	case OPCODE_CMP:
-	case OPCODE_MAD:
-		inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
-		inst->SrcReg[1] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[1]);
-		inst->SrcReg[2] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[2]);
-		break;
-	default:
-		// Scalar instructions needn't be reswizzled
-		break;
-	}
-}
-
 static void unalias_srcregs(struct prog_instruction *inst, GLuint oldindex, GLuint newindex)
 {
 	int nsrc = _mesa_num_inst_src_regs(inst->Opcode);
@@ -189,11 +155,6 @@ static void process_instruction(struct nqssadce_state* s)
 		return;
 
 	if (inst->Opcode != OPCODE_KIL) {
-		if (s->Descr->RewriteDepthOut) {
-			if (inst->DstReg.File == PROGRAM_OUTPUT && inst->DstReg.Index == FRAG_RESULT_DEPTH)
-				rewrite_depth_out(inst);
-		}
-
 		struct register_state *regstate = get_reg_state(s, inst->DstReg.File, inst->DstReg.Index);
 		if (!regstate) {
 			_mesa_problem(s->Ctx, "NqssaDce: bad destination register (%i[%i])\n",
diff --git a/src/mesa/drivers/dri/r300/radeon_nqssadce.h b/src/mesa/drivers/dri/r300/radeon_nqssadce.h
index a4f94abcb6..e3341692e4 100644
--- a/src/mesa/drivers/dri/r300/radeon_nqssadce.h
+++ b/src/mesa/drivers/dri/r300/radeon_nqssadce.h
@@ -83,14 +83,10 @@ struct radeon_nqssadce_descr {
 	 */
 	void (*BuildSwizzle)(struct nqssadce_state*, struct prog_dst_register dst, struct prog_src_register src);
 
-	/**
-	 * Rewrite instructions that write to DEPR.z to write to DEPR.w
-	 * instead (rewriting is done *before* the WriteMask test).
-	 */
-	GLboolean RewriteDepthOut;
 	void *Data;
 };
 
 void radeonNqssaDce(GLcontext *ctx, struct gl_program *p, struct radeon_nqssadce_descr* descr);
+struct prog_src_register lmul_swizzle(GLuint swizzle, struct prog_src_register srcreg);
 
 #endif /* __RADEON_PROGRAM_NQSSADCE_H_ */
-- 
cgit v1.2.3


From 96b2eb18c5059d441873bfa562bd2ff9dec46a51 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 4 Jul 2009 15:22:22 +0200
Subject: r300: handle ARB_vertex_program specific instructions in NQSSADCE

---
 src/mesa/drivers/dri/r300/radeon_nqssadce.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/radeon_nqssadce.c b/src/mesa/drivers/dri/r300/radeon_nqssadce.c
index 82dfc31dd0..c56598b4a5 100644
--- a/src/mesa/drivers/dri/r300/radeon_nqssadce.c
+++ b/src/mesa/drivers/dri/r300/radeon_nqssadce.c
@@ -188,6 +188,8 @@ static void process_instruction(struct nqssadce_state* s)
 	case OPCODE_MAX:
 	case OPCODE_MIN:
 	case OPCODE_MUL:
+	case OPCODE_SGE:
+	case OPCODE_SLT:
 		inst = track_used_srcreg(s, inst, 0, inst->DstReg.WriteMask);
 		inst = track_used_srcreg(s, inst, 1, inst->DstReg.WriteMask);
 		break;
@@ -219,6 +221,18 @@ static void process_instruction(struct nqssadce_state* s)
 	case OPCODE_TXP:
 		inst = track_used_srcreg(s, inst, 0, 0xf);
 		break;
+	case OPCODE_DST:
+		inst = track_used_srcreg(s, inst, 0, 0x6);
+		inst = track_used_srcreg(s, inst, 1, 0xa);
+		break;
+	case OPCODE_EXP:
+	case OPCODE_LOG:
+	case OPCODE_POW:
+		inst = track_used_srcreg(s, inst, 0, 0x3);
+		break;
+	case OPCODE_LIT:
+		inst = track_used_srcreg(s, inst, 0, 0xb);
+		break;
 	default:
 		_mesa_problem(s->Ctx, "NqssaDce: Unknown opcode %d\n", inst->Opcode);
 		return;
-- 
cgit v1.2.3


From 12a6d73c7590c37ec8ae3f2c8c737791e4461d77 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Mon, 13 Jul 2009 19:23:18 +0200
Subject: r300: handle relative addressing in NQSSADCE

---
 src/mesa/drivers/dri/r300/radeon_nqssadce.c | 10 +++++++++-
 src/mesa/drivers/dri/r300/radeon_nqssadce.h |  1 +
 2 files changed, 10 insertions(+), 1 deletion(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/radeon_nqssadce.c b/src/mesa/drivers/dri/r300/radeon_nqssadce.c
index c56598b4a5..d917da58bb 100644
--- a/src/mesa/drivers/dri/r300/radeon_nqssadce.c
+++ b/src/mesa/drivers/dri/r300/radeon_nqssadce.c
@@ -46,6 +46,7 @@ static struct register_state *get_reg_state(struct nqssadce_state* s, GLuint fil
 	switch(file) {
 	case PROGRAM_TEMPORARY: return &s->Temps[index];
 	case PROGRAM_OUTPUT: return &s->Outputs[index];
+	case PROGRAM_ADDRESS: return &s->Address;
 	default: return 0;
 	}
 }
@@ -114,7 +115,13 @@ static struct prog_instruction* track_used_srcreg(struct nqssadce_state* s,
 		deswz_source = sourced;
 	}
 
-	struct register_state *regstate = get_reg_state(s, inst->SrcReg[src].File, inst->SrcReg[src].Index);
+	struct register_state *regstate;
+
+	if (inst->SrcReg[src].RelAddr)
+		regstate = get_reg_state(s, PROGRAM_ADDRESS, 0);
+	else
+		regstate = get_reg_state(s, inst->SrcReg[src].File, inst->SrcReg[src].Index);
+
 	if (regstate)
 		regstate->Sourced |= deswz_source & 0xf;
 
@@ -178,6 +185,7 @@ static void process_instruction(struct nqssadce_state* s)
 	 * might change the instruction stream under us, so we have
 	 * to be careful with the inst pointer. */
 	switch (inst->Opcode) {
+	case OPCODE_ARL:
 	case OPCODE_DDX:
 	case OPCODE_DDY:
 	case OPCODE_FRC:
diff --git a/src/mesa/drivers/dri/r300/radeon_nqssadce.h b/src/mesa/drivers/dri/r300/radeon_nqssadce.h
index e3341692e4..8626f21c25 100644
--- a/src/mesa/drivers/dri/r300/radeon_nqssadce.h
+++ b/src/mesa/drivers/dri/r300/radeon_nqssadce.h
@@ -58,6 +58,7 @@ struct nqssadce_state {
 	 */
 	struct register_state Temps[MAX_PROGRAM_TEMPS];
 	struct register_state Outputs[VERT_RESULT_MAX];
+	struct register_state Address;
 };
 
 
-- 
cgit v1.2.3


From 70448b9f95b4ca56526458d207a28727f71e8d3c Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 4 Jul 2009 16:52:48 +0200
Subject: r300: operate on copy of a program when pairing instructions

We need to keep unpaired program for vertex program NQSSADCE.
---
 src/mesa/drivers/dri/r300/radeon_program_pair.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/radeon_program_pair.c b/src/mesa/drivers/dri/r300/radeon_program_pair.c
index 906d36e522..d6fb474cf2 100644
--- a/src/mesa/drivers/dri/r300/radeon_program_pair.c
+++ b/src/mesa/drivers/dri/r300/radeon_program_pair.c
@@ -870,7 +870,7 @@ GLboolean radeonPairProgram(GLcontext *ctx, struct gl_program *program,
 
 	_mesa_bzero(&s, sizeof(s));
 	s.Ctx = ctx;
-	s.Program = program;
+	s.Program = _mesa_clone_program(ctx, program);
 	s.Handler = handler;
 	s.UserData = userdata;
 	s.Debug = (RADEON_DEBUG & DEBUG_PIXEL) ? GL_TRUE : GL_FALSE;
@@ -904,6 +904,8 @@ GLboolean radeonPairProgram(GLcontext *ctx, struct gl_program *program,
 	_mesa_free(s.ValuePool);
 	_mesa_free(s.ReaderPool);
 
+	_mesa_reference_program(ctx, &s.Program, NULL);
+
 	return !s.Error;
 }
 
-- 
cgit v1.2.3


From 65d9f23c7c3e99dfe038b1306f96930e1228de11 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sun, 5 Jul 2009 02:34:48 +0200
Subject: r300: use NQSSADCE for vertex programs

---
 src/mesa/drivers/dri/r300/r300_vertprog.c | 133 +++++++++++++++++++++++++++++-
 1 file changed, 132 insertions(+), 1 deletion(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index 6f9b4db40f..fb94eea07a 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -34,11 +34,13 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "shader/program.h"
 #include "shader/programopt.h"
 #include "shader/prog_instruction.h"
+#include "shader/prog_optimize.h"
 #include "shader/prog_parameter.h"
 #include "shader/prog_print.h"
 #include "shader/prog_statevars.h"
 #include "tnl/tnl.h"
 
+#include "radeon_nqssadce.h"
 #include "r300_context.h"
 #include "r300_state.h"
 
@@ -1537,6 +1539,86 @@ static void addArtificialOutputs(GLcontext *ctx, struct gl_program *prog)
 
 #undef ADD_OUTPUT
 
+static GLuint getUsedComponents(const GLuint swizzle)
+{
+	GLuint ret;
+
+	ret = 0;
+
+	/* need to mask out ZERO, ONE and NIL swizzles */
+	ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_X) & 0x3);
+	ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_Y) & 0x3);
+	ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_Z) & 0x3);
+	ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_W) & 0x3);
+
+	return ret;
+}
+
+static GLuint trackUsedComponents(const struct gl_program *prog, const GLuint attrib)
+{
+	struct prog_instruction *inst;
+	int tmp, i;
+	GLuint ret;
+
+	inst = prog->Instructions;
+	ret = 0;
+	while (inst->Opcode != OPCODE_END) {
+		tmp = _mesa_num_inst_src_regs(inst->Opcode);
+		for (i = 0; i < tmp; ++i) {
+			if (inst->SrcReg[i].File == PROGRAM_INPUT && inst->SrcReg[i].Index == attrib) {
+				ret |= getUsedComponents(inst->SrcReg[i].Swizzle);
+			}
+		}
+		++inst;
+	}
+
+	return ret;
+}
+
+static void nqssadceInit(struct nqssadce_state* s)
+{
+	r300ContextPtr r300 = R300_CONTEXT(s->Ctx);
+	GLuint fp_reads;
+
+	fp_reads = r300->selected_fp->Base->InputsRead;
+	{
+		GLuint tmp;
+
+		if (fp_reads & FRAG_BIT_COL0) {
+				tmp = trackUsedComponents(r300->selected_fp->Base, FRAG_ATTRIB_COL0);
+				s->Outputs[VERT_RESULT_COL0].Sourced = tmp;
+				s->Outputs[VERT_RESULT_BFC0].Sourced = tmp;
+		}
+
+		if (fp_reads & FRAG_BIT_COL1) {
+				tmp = trackUsedComponents(r300->selected_fp->Base, FRAG_ATTRIB_COL1);
+				s->Outputs[VERT_RESULT_COL1].Sourced = tmp;
+				s->Outputs[VERT_RESULT_BFC1].Sourced = tmp;
+		}
+	}
+
+	{
+		int i;
+		for (i = 0; i < 8; ++i) {
+			if (fp_reads & FRAG_BIT_TEX(i)) {
+				s->Outputs[VERT_RESULT_TEX0 + i].Sourced = trackUsedComponents(r300->selected_fp->Base, FRAG_ATTRIB_TEX0 + i);
+			}
+		}
+	}
+
+	s->Outputs[VERT_RESULT_HPOS].Sourced = WRITEMASK_XYZW;
+	if (s->Program->OutputsWritten & (1 << VERT_RESULT_PSIZ))
+		s->Outputs[VERT_RESULT_PSIZ].Sourced = WRITEMASK_X;
+}
+
+static GLboolean swizzleIsNative(GLuint opcode, struct prog_src_register reg)
+{
+	(void) opcode;
+	(void) reg;
+
+	return GL_TRUE;
+}
+
 static struct r300_vertex_program *build_program(GLcontext *ctx,
 						 struct r300_vertex_program_key *wanted_key,
 						 const struct gl_vertex_program *mesa_vp)
@@ -1579,8 +1661,57 @@ static struct r300_vertex_program *build_program(GLcontext *ctx,
 		fflush(stdout);
 	}
 
+	{
+		struct radeon_nqssadce_descr nqssadce = {
+			.Init = &nqssadceInit,
+			.IsNativeSwizzle = &swizzleIsNative,
+			.BuildSwizzle = NULL
+		};
+		radeonNqssaDce(ctx, prog, &nqssadce);
+
+		/* We need this step for reusing temporary registers */
+		_mesa_optimize_program(ctx, prog);
+
+		if (RADEON_DEBUG & DEBUG_VERTS) {
+			fprintf(stderr, "Vertex program after NQSSADCE:\n");
+			_mesa_print_program(prog);
+			fflush(stdout);
+		}
+	}
+
 	assert(prog->NumInstructions);
-	vp->num_temporaries = prog->NumTemporaries;
+	{
+		struct prog_instruction *inst;
+		int max, i, tmp;
+
+		inst = prog->Instructions;
+		max = -1;
+		while (inst->Opcode != OPCODE_END) {
+			tmp = _mesa_num_inst_src_regs(inst->Opcode);
+			for (i = 0; i < tmp; ++i) {
+				if (inst->SrcReg[i].File == PROGRAM_TEMPORARY) {
+					if ((int) inst->SrcReg[i].Index > max) {
+						max = inst->SrcReg[i].Index;
+					}
+				}
+			}
+
+			if (_mesa_num_inst_dst_regs(inst->Opcode)) {
+				if (inst->DstReg.File == PROGRAM_TEMPORARY) {
+					if ((int) inst->DstReg.Index > max) {
+						max = inst->DstReg.Index;
+					}
+				}
+			}
+			++inst;
+		}
+
+		/* We actually want highest index of used temporary register,
+		 * not the number of temporaries used.
+		 * These values aren't always the same.
+		 */
+		vp->num_temporaries = max + 1;
+	}
 
 	return vp;
 }
-- 
cgit v1.2.3


From bdc8a95fc906dcc617995479fbffbf8fdd1f4e34 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 4 Jul 2009 23:38:59 +0200
Subject: r300: removed unnecessary params

We don't have check which attributes are used by fragment program - it's already done by NQSSADCE.
---
 src/mesa/drivers/dri/r300/r300_emit.c  | 10 +++++-----
 src/mesa/drivers/dri/r300/r300_emit.h  |  4 ++--
 src/mesa/drivers/dri/r300/r300_state.c |  4 ++--
 3 files changed, 9 insertions(+), 9 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c
index 707d1284ed..feb3370f37 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.c
+++ b/src/mesa/drivers/dri/r300/r300_emit.c
@@ -81,17 +81,17 @@ GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead)
 	return vic_1;
 }
 
-GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint vp_writes, GLuint fp_reads)
+GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint vp_writes)
 {
 	GLuint ret = 0;
 
 	if (vp_writes & (1 << VERT_RESULT_HPOS))
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT;
 
-	if (vp_writes & (1 << VERT_RESULT_COL0) && fp_reads & FRAG_BIT_COL0)
+	if (vp_writes & (1 << VERT_RESULT_COL0))
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT;
 
-	if (vp_writes & (1 << VERT_RESULT_COL1) && fp_reads & FRAG_BIT_COL1)
+	if (vp_writes & (1 << VERT_RESULT_COL1))
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT;
 
 	/* Two sided lighting works only if all 4 colors are written */
@@ -105,12 +105,12 @@ GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint vp_writes, GLuint fp_reads)
 	return ret;
 }
 
-GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint vp_writes, GLuint fp_reads)
+GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint vp_writes)
 {
 	GLuint i, ret = 0, first_free_texcoord = 0;
 
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-		if (vp_writes & (1 << (VERT_RESULT_TEX0 + i)) && fp_reads & FRAG_BIT_TEX(i)) {
+		if (vp_writes & (1 << (VERT_RESULT_TEX0 + i))) {
 			ret |= (4 << (3 * first_free_texcoord));
 			++first_free_texcoord;
 		}
diff --git a/src/mesa/drivers/dri/r300/r300_emit.h b/src/mesa/drivers/dri/r300/r300_emit.h
index 2fb8b82d3a..3f8c60ffae 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.h
+++ b/src/mesa/drivers/dri/r300/r300_emit.h
@@ -225,7 +225,7 @@ extern void r300EmitCacheFlush(r300ContextPtr rmesa);
 
 extern GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead);
 extern GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead);
-extern GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint vp_writes, GLuint fp_reads);
-extern GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint vp_writes, GLuint fp_reads);
+extern GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint vp_writes);
+extern GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint vp_writes);
 
 #endif
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index bae2f0f3cb..4d504d14e3 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -2225,8 +2225,8 @@ void r300SetupVAP(GLcontext *ctx, GLuint InputsRead, GLuint OutputsWritten)
 
 	rmesa->hw.vic.cmd[R300_VIC_CNTL_0] = r300VAPInputCntl0(ctx, InputsRead);
 	rmesa->hw.vic.cmd[R300_VIC_CNTL_1] = r300VAPInputCntl1(ctx, InputsRead);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_0] = r300VAPOutputCntl0(ctx, OutputsWritten, rmesa->selected_fp->Base->InputsRead);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = r300VAPOutputCntl1(ctx, OutputsWritten, rmesa->selected_fp->Base->InputsRead);
+	rmesa->hw.vof.cmd[R300_VOF_CNTL_0] = r300VAPOutputCntl0(ctx, OutputsWritten);
+	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = r300VAPOutputCntl1(ctx, OutputsWritten);
 }
 
 void r300UpdateShaderStates(r300ContextPtr rmesa)
-- 
cgit v1.2.3


From 4efb9f053c84fbdced373cc62fe06b3158b59015 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sun, 5 Jul 2009 03:17:32 +0200
Subject: r300: fix WPOS for SWTCL

---
 src/mesa/drivers/dri/r300/r300_swtcl.c | 30 ++++++++++++++++--------------
 1 file changed, 16 insertions(+), 14 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c
index d73508d36e..56ed519cf4 100644
--- a/src/mesa/drivers/dri/r300/r300_swtcl.c
+++ b/src/mesa/drivers/dri/r300/r300_swtcl.c
@@ -150,6 +150,22 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead,  GLuint *_
 		ADD_ATTR(VERT_ATTRIB_POINT_SIZE, R300_DATA_TYPE_FLOAT_1, SWTCL_OVM_POINT_SIZE, swiz, MASK_X, 0);
 	}
 
+	if (rmesa->selected_fp->wpos_attr != FRAG_ATTRIB_MAX) {
+		int tex_id = rmesa->selected_fp->wpos_attr - FRAG_ATTRIB_TEX0;
+
+		VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_POS];
+		VB->TexCoordPtr[tex_id] = VB->AttribPtr[VERT_ATTRIB_POS];
+		RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id);
+	}
+
+	if (rmesa->selected_fp->fog_attr != FRAG_ATTRIB_MAX) {
+		int tex_id = rmesa->selected_fp->fog_attr - FRAG_ATTRIB_TEX0;
+
+		VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG];
+		VB->TexCoordPtr[tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG];
+		RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id);
+	}
+
 	/**
 	 *  Sending only one texcoord component may lead to lock up,
 	 *  so for all textures always output 4 texcoord components to RS.
@@ -192,20 +208,6 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead,  GLuint *_
 		}
 	}
 
-	if (rmesa->selected_fp->wpos_attr != FRAG_ATTRIB_MAX) {
-		int tex_id = rmesa->selected_fp->wpos_attr - FRAG_ATTRIB_TEX0;
-
-		VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_POS];
-		RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id);
-	}
-
-	if (rmesa->selected_fp->fog_attr != FRAG_ATTRIB_MAX) {
-		int tex_id = rmesa->selected_fp->fog_attr - FRAG_ATTRIB_TEX0;
-
-		VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG];
-		RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id);
-	}
-
 	if (first_free_tex >= ctx->Const.MaxTextureUnits) {
 		fprintf(stderr, "\tout of free texcoords to write fog coordinate\n");
 		_mesa_exit(-1);
-- 
cgit v1.2.3


From 4eff323731b0d65e1f2817e96435807418d833cc Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 11 Jul 2009 15:09:20 +0200
Subject: r300: hw can handle per component negations in vertex shaders

Reported-by: Nicolai Haehnle <nhaehnle@gmail.com>
---
 src/mesa/drivers/dri/r300/r300_vertprog.c | 27 ++-------------------------
 1 file changed, 2 insertions(+), 25 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index fb94eea07a..ea8d25d5a9 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -1388,32 +1388,9 @@ static int translateSUB(struct gl_program *prog, int pos)
 
 static int translateSWZ(struct gl_program *prog, int pos)
 {
-	struct prog_instruction *inst;
-	GLuint orig_negate, orig_writemask;
-
-	inst = &prog->Instructions[pos];
-	orig_negate = inst->SrcReg[0].Negate;
-	orig_writemask = inst->DstReg.WriteMask;
-
-	inst->Opcode = OPCODE_MOV;
+	prog->Instructions[pos].Opcode = OPCODE_MOV;
 
-	/* If all relevant components are either negated or not negated at the same time, we are ok.
-	 */
-	if ((orig_negate & orig_writemask) == 0 || (orig_negate & orig_writemask) == (NEGATE_XYZW & orig_writemask))
-		return 0;
-
-	_mesa_insert_instructions(prog, pos + 1, 1);
-
-	inst = &prog->Instructions[pos];
-	inst->DstReg.WriteMask = orig_writemask & (orig_negate ^ NEGATE_XYZW);
-	inst->SrcReg[0].Negate = NEGATE_NONE;
-	++inst;
-
-	*inst = *(inst-1);
-	inst->DstReg.WriteMask = orig_writemask & orig_negate;
-	inst->SrcReg[0].Negate = NEGATE_XYZW;
-
-	return 1;
+	return 0;
 }
 
 static int translateXPD(struct gl_program *prog, int pos)
-- 
cgit v1.2.3


From 48cc352a71c728e0ce092dd7d7ec0945db304907 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 11 Jul 2009 15:53:01 +0200
Subject: r300: fix StrideB == 0 case when converting data format
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Reported-by: Nicolai Hähnle <nhaehnle@gmail.com>
---
 src/mesa/drivers/dri/r300/r300_draw.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c
index 5420293b91..9769ff5399 100644
--- a/src/mesa/drivers/dri/r300/r300_draw.c
+++ b/src/mesa/drivers/dri/r300/r300_draw.c
@@ -195,6 +195,11 @@ static void r300TranslateAttrib(GLcontext *ctx, GLuint attr, int count, const st
 		}
 
 		GLfloat *dst_ptr, *tmp;
+
+		/* Convert value for first element only */
+		if (input->StrideB == 0)
+			count = 1;
+
 		tmp = dst_ptr = _mesa_malloc(sizeof(GLfloat) * input->Size * count);
 
 		switch (input->Type) {
@@ -228,7 +233,11 @@ static void r300TranslateAttrib(GLcontext *ctx, GLuint attr, int count, const st
 		type = GL_FLOAT;
 		r300_attr.free_needed = GL_TRUE;
 		r300_attr.data = tmp;
-		r300_attr.stride = sizeof(GLfloat) * input->Size;
+		if (input->StrideB == 0) {
+			r300_attr.stride = 0;
+		} else {
+			r300_attr.stride = sizeof(GLfloat) * input->Size;
+		}
 		r300_attr.dwords = input->Size;
 	} else {
 		type = input->Type;
-- 
cgit v1.2.3


From ec854729d12a533ed00b9a9e6a5942aede2f5d1e Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 11 Jul 2009 16:16:11 +0200
Subject: r300: fix indentation
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Reported-by: Nicolai Hähnle <nhaehnle@gmail.com>
---
 src/mesa/drivers/dri/r300/r300_fragprog_common.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.c b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
index d5ef18b2ad..c7914649dd 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_common.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
@@ -368,10 +368,10 @@ struct r300_fragment_program *r300SelectFragmentShader(GLcontext *ctx)
 
 	fp = fp_list->progs;
 	while (fp) {
-	       if (_mesa_memcmp(&fp->state, &state, sizeof(state)) == 0) {
-	               return r300->selected_fp = fp;
-	       }
-	       fp = fp->next;
+		if (_mesa_memcmp(&fp->state, &state, sizeof(state)) == 0) {
+			return r300->selected_fp = fp;
+		}
+		fp = fp->next;
 	}
 
 	fp = _mesa_calloc(sizeof(struct r300_fragment_program));
-- 
cgit v1.2.3


From a0204ce456435f6ee38d8c14d25ae251940bf55f Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 11 Jul 2009 16:26:23 +0200
Subject: r300: document r300_fragment_program_cont struct

---
 src/mesa/drivers/dri/r300/r300_context.h | 5 +++++
 1 file changed, 5 insertions(+)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 32104b8158..f40e8ac394 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -569,7 +569,12 @@ struct r300_fragment_program {
 };
 
 struct r300_fragment_program_cont {
+	/* This is the unmodified fragment program mesa provided us with.
+	 * We need to keep it unchanged because we may need to create another
+	 * hw specific fragment program based on this
+	 */
 	struct gl_fragment_program Base;
+	/* This is the list of hw specific fragment programs derived from Base */
 	struct r300_fragment_program *progs;
 };
 
-- 
cgit v1.2.3


From b3716eeb61b75dc661a1485f9bea9275a88a1d81 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 11 Jul 2009 16:34:04 +0200
Subject: r300: document r300_vertex_program_cont structure
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Reported-by: Nicolai Hähnle <nhaehnle@gmail.com>
---
 src/mesa/drivers/dri/r300/r300_context.h | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index f40e8ac394..f7af7d4e57 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -432,7 +432,12 @@ struct r300_vertex_program {
 };
 
 struct r300_vertex_program_cont {
-	struct gl_vertex_program mesa_program;	/* Must be first */
+	/* This is the unmodified vertex program mesa provided us with.
+	 * We need to keep it unchanged because we may need to create another
+	 * hw specific vertex program based on this.
+	 */
+	struct gl_vertex_program mesa_program;
+	/* This is the list of hw specific vertex programs derived from mesa_program */
 	struct r300_vertex_program *progs;
 };
 
@@ -571,7 +576,7 @@ struct r300_fragment_program {
 struct r300_fragment_program_cont {
 	/* This is the unmodified fragment program mesa provided us with.
 	 * We need to keep it unchanged because we may need to create another
-	 * hw specific fragment program based on this
+	 * hw specific fragment program based on this.
 	 */
 	struct gl_fragment_program Base;
 	/* This is the list of hw specific fragment programs derived from Base */
-- 
cgit v1.2.3


From 1a5520fcd3842cc3198bff143d2af5c169eebc26 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 11 Jul 2009 16:37:11 +0200
Subject: r300: move variables declarations

---
 src/mesa/drivers/dri/r300/radeon_nqssadce.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/radeon_nqssadce.c b/src/mesa/drivers/dri/r300/radeon_nqssadce.c
index d917da58bb..840c9733b1 100644
--- a/src/mesa/drivers/dri/r300/radeon_nqssadce.c
+++ b/src/mesa/drivers/dri/r300/radeon_nqssadce.c
@@ -250,7 +250,6 @@ static void process_instruction(struct nqssadce_state* s)
 static void calculateInputsOutputs(struct gl_program *p)
 {
 	struct prog_instruction *inst;
-	int i, tmp;
 	GLuint InputsRead, OutputsWritten;
 
 	inst = p->Instructions;
@@ -258,8 +257,10 @@ static void calculateInputsOutputs(struct gl_program *p)
 	OutputsWritten = 0;
 	while (inst->Opcode != OPCODE_END)
 	{
-		tmp = _mesa_num_inst_src_regs(inst->Opcode);
-		for (i = 0; i < tmp; ++i) {
+		int i, num_src_regs;
+
+		num_src_regs = _mesa_num_inst_src_regs(inst->Opcode);
+		for (i = 0; i < num_src_regs; ++i) {
 			if (inst->SrcReg[i].File == PROGRAM_INPUT)
 				InputsRead |= 1 << inst->SrcReg[i].Index;
 		}
-- 
cgit v1.2.3


From acd33600411102872af579edc4206b61eb51bb65 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 11 Jul 2009 16:41:52 +0200
Subject: r300: minor fix
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Split initializations becase the vars are of different type.

Reported-by: Nicolai Hähnle <nhaehnle@gmail.com>
---
 src/mesa/drivers/dri/r300/r300_vertprog.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index ea8d25d5a9..703fba4874 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -1473,7 +1473,8 @@ static void addArtificialOutputs(GLcontext *ctx, struct gl_program *prog)
 	GLuint OutputsAdded, FpReads;
 	int i, count;
 
-	OutputsAdded = count = 0;
+	OutputsAdded = 0;
+	count = 0;
 	FpReads = r300->selected_fp->Base->InputsRead;
 
 	ADD_OUTPUT(FRAG_ATTRIB_COL0, VERT_RESULT_COL0);
-- 
cgit v1.2.3


From 3f5382819e31071c2af6cb39c1ca09bf3243f83f Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 11 Jul 2009 19:10:58 +0200
Subject: r300: fix swizzle masking in getUsedComponents

---
 src/mesa/drivers/dri/r300/r300_vertprog.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index 703fba4874..9f807dbc99 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -1524,10 +1524,14 @@ static GLuint getUsedComponents(const GLuint swizzle)
 	ret = 0;
 
 	/* need to mask out ZERO, ONE and NIL swizzles */
-	ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_X) & 0x3);
-	ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_Y) & 0x3);
-	ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_Z) & 0x3);
-	ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_W) & 0x3);
+	if (GET_SWZ(swizzle, SWIZZLE_X) <= SWIZZLE_W)
+		ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_X));
+	if (GET_SWZ(swizzle, SWIZZLE_Y) <= SWIZZLE_W)
+		ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_Y));
+	if (GET_SWZ(swizzle, SWIZZLE_Z) <= SWIZZLE_W)
+		ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_Z));
+	if (GET_SWZ(swizzle, SWIZZLE_W) <= SWIZZLE_W)
+		ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_W));
 
 	return ret;
 }
-- 
cgit v1.2.3


From f06910f6c34f199dc187bd69ff1f6889ba498217 Mon Sep 17 00:00:00 2001
From: Nicolai Hähnle <nhaehnle@gmail.com>
Date: Sun, 12 Jul 2009 02:03:25 +0000
Subject: r300: Fix fogcoord rewriting
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

We only care about the actual fogcoord itself now, reducing the rewriting
done for the vertex program.

The rewriting of source operand swizzles in the fragment program takes
care that fogcoord.yzw = 001.

This should fix fogcoord rewriting entirely, which had been horribly
broken in the face of dot-product instructions, and just broken (though
not horribly so) in the face of almost every other instruction (the W
component would be incorrect for most arithmetic instructions).

Signed-off-by: Nicolai Hähnle <nhaehnle@gmail.com>
---
 src/mesa/drivers/dri/r300/r300_fragprog_common.c | 14 ++++-
 src/mesa/drivers/dri/r300/r300_vertprog.c        | 65 ++++++++----------------
 src/mesa/drivers/dri/r300/radeon_program.h       | 32 ++++++++++++
 3 files changed, 66 insertions(+), 45 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.c b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
index c7914649dd..f5c4c0f4a0 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_common.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
@@ -162,6 +162,14 @@ static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler)
 	}
 }
 
+
+/**
+ * Rewrite fragment.fogcoord to use a texture coordinate slot.
+ * Note that fogcoord is forced into an X001 pattern, and this enforcement
+ * is done here.
+ *
+ * See also the counterpart rewriting for vertex programs.
+ */
 static void rewriteFog(struct r300_fragment_program_compiler *compiler)
 {
 	struct r300_fragment_program *fp = compiler->fp;
@@ -193,8 +201,12 @@ static void rewriteFog(struct r300_fragment_program_compiler *compiler)
 		while (inst->Opcode != OPCODE_END) {
 			const int src_regs = _mesa_num_inst_src_regs(inst->Opcode);
 			for (i = 0; i < src_regs; ++i) {
-				if (inst->SrcReg[i].File == PROGRAM_INPUT && inst->SrcReg[i].Index == FRAG_ATTRIB_FOGC)
+				if (inst->SrcReg[i].File == PROGRAM_INPUT && inst->SrcReg[i].Index == FRAG_ATTRIB_FOGC) {
 					inst->SrcReg[i].Index = fp->fog_attr;
+					inst->SrcReg[i].Swizzle = combine_swizzles(
+						MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE),
+						inst->SrcReg[i].Swizzle);
+				}
 			}
 			++inst;
 		}
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index 9f807dbc99..c0b116156a 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -74,33 +74,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 		u_temp_i=VSF_MAX_FRAGMENT_TEMPS-1; \
 	} while (0)
 
-static GLuint combineSwizzles(GLuint src, GLuint swz_x, GLuint swz_y, GLuint swz_z, GLuint swz_w)
-{
-	GLuint ret = 0;
-
-	if (swz_x == SWIZZLE_ZERO || swz_x == SWIZZLE_ONE)
-		ret |= swz_x;
-	else
-		ret |= GET_SWZ(src, swz_x);
-
-	if (swz_y == SWIZZLE_ZERO || swz_y == SWIZZLE_ONE)
-		ret |= swz_y << 3;
-	else
-		ret |= GET_SWZ(src, swz_y) << 3;
-
-	if (swz_z == SWIZZLE_ZERO || swz_z == SWIZZLE_ONE)
-		ret |= swz_z << 6;
-	else
-		ret |= GET_SWZ(src, swz_z) << 6;
-
-	if (swz_w == SWIZZLE_ZERO || swz_w == SWIZZLE_ONE)
-		ret |= swz_w << 9;
-	else
-		ret |= GET_SWZ(src, swz_w) << 9;
-
-	return ret;
-}
-
 static int r300VertexProgUpdateParams(GLcontext * ctx, struct gl_vertex_program *vp, float *dst)
 {
 	int pi;
@@ -1282,24 +1255,28 @@ static void pos_as_texcoord(struct gl_program *prog, int tex_id)
 }
 
 /**
- @TODO
-  We can put X001 swizzle only if input components are directly mapped from output components.
-  For some insts we need to skip source swizzles and add: MOV OUTPUT[fog_attr].yzw, CONST[0].0001
+ * The fogcoord attribute is special in that only the first component
+ * is relevant, and the remaining components are always fixed (when read
+ * from by the fragment program) to yield an X001 pattern.
+ *
+ * We need to enforce this either in the vertex program or in the fragment
+ * program, and this code chooses not to enforce it in the vertex program.
+ * This is slightly cheaper, as long as the fragment program does not use
+ * weird swizzles.
+ *
+ * And it seems that usually, weird swizzles are not used, so...
+ *
+ * See also the counterpart rewriting for fragment programs.
  */
 static void fog_as_texcoord(struct gl_program *prog, int tex_id)
 {
 	struct prog_instruction *vpi;
-	int i;
 
 	vpi = prog->Instructions;
 	while (vpi->Opcode != OPCODE_END) {
 		if (vpi->DstReg.File == PROGRAM_OUTPUT && vpi->DstReg.Index == VERT_RESULT_FOGC) {
 			vpi->DstReg.Index = VERT_RESULT_TEX0 + tex_id;
-			vpi->DstReg.WriteMask = WRITEMASK_XYZW;
-
-			for (i = 0; i < _mesa_num_inst_src_regs(vpi->Opcode); ++i) {
-				vpi->SrcReg[i].Swizzle = combineSwizzles(vpi->SrcReg[i].Swizzle, SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE);
-			}
+			vpi->DstReg.WriteMask = WRITEMASK_X;
 		}
 
 		++vpi;
@@ -1329,7 +1306,7 @@ static int translateDP3(struct gl_program *prog, int pos)
 	inst = &prog->Instructions[pos];
 
 	inst->Opcode = OPCODE_DP4;
-	inst->SrcReg[0].Swizzle = combineSwizzles(inst->SrcReg[0].Swizzle, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
+	inst->SrcReg[0].Swizzle = combine_swizzles4(inst->SrcReg[0].Swizzle, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
 
 	return 0;
 }
@@ -1341,7 +1318,7 @@ static int translateDPH(struct gl_program *prog, int pos)
 	inst = &prog->Instructions[pos];
 
 	inst->Opcode = OPCODE_DP4;
-	inst->SrcReg[0].Swizzle = combineSwizzles(inst->SrcReg[0].Swizzle, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
+	inst->SrcReg[0].Swizzle = combine_swizzles4(inst->SrcReg[0].Swizzle, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
 
 	return 0;
 }
@@ -1409,13 +1386,13 @@ static int translateXPD(struct gl_program *prog, int pos)
 	inst->Opcode = OPCODE_MUL;
 	inst->DstReg.File = PROGRAM_TEMPORARY;
 	inst->DstReg.Index = tmp_idx;
-	inst->SrcReg[0].Swizzle = combineSwizzles(inst->SrcReg[0].Swizzle, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_W);
-	inst->SrcReg[1].Swizzle = combineSwizzles(inst->SrcReg[1].Swizzle, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W);
+	inst->SrcReg[0].Swizzle = combine_swizzles4(inst->SrcReg[0].Swizzle, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_W);
+	inst->SrcReg[1].Swizzle = combine_swizzles4(inst->SrcReg[1].Swizzle, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W);
 	++inst;
 
 	inst->Opcode = OPCODE_MAD;
-	inst->SrcReg[0].Swizzle = combineSwizzles(inst->SrcReg[0].Swizzle, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W);
-	inst->SrcReg[1].Swizzle = combineSwizzles(inst->SrcReg[1].Swizzle, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_W);
+	inst->SrcReg[0].Swizzle = combine_swizzles4(inst->SrcReg[0].Swizzle, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W);
+	inst->SrcReg[1].Swizzle = combine_swizzles4(inst->SrcReg[1].Swizzle, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_W);
 	inst->SrcReg[1].Negate ^= NEGATE_XYZW;
 	inst->SrcReg[2].File = PROGRAM_TEMPORARY;
 	inst->SrcReg[2].Index = tmp_idx;
@@ -1768,12 +1745,12 @@ void r300SetupVertexProgram(r300ContextPtr rmesa)
 	struct r300_vertex_program *prog = rmesa->selected_vp;
 	int inst_count = 0;
 	int param_count = 0;
-	
+
 	/* Reset state, in case we don't use something */
 	((drm_r300_cmd_header_t *) rmesa->hw.vpp.cmd)->vpu.count = 0;
 	((drm_r300_cmd_header_t *) rmesa->hw.vpi.cmd)->vpu.count = 0;
 	((drm_r300_cmd_header_t *) rmesa->hw.vps.cmd)->vpu.count = 0;
-	
+
 	R300_STATECHANGE(rmesa, vpp);
 	param_count = r300VertexProgUpdateParams(ctx, prog->Base, (float *)&rmesa->hw.vpp.cmd[R300_VPP_PARAM_0]);
 	bump_vpu_count(rmesa->hw.vpp.cmd, param_count);
diff --git a/src/mesa/drivers/dri/r300/radeon_program.h b/src/mesa/drivers/dri/r300/radeon_program.h
index b411f69bc8..88474d43a2 100644
--- a/src/mesa/drivers/dri/r300/radeon_program.h
+++ b/src/mesa/drivers/dri/r300/radeon_program.h
@@ -52,6 +52,38 @@ enum {
 #define SWIZZLE_0000 MAKE_SWIZZLE4(SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO)
 #define SWIZZLE_1111 MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE)
 
+static inline GLuint get_swz(GLuint swz, GLuint idx)
+{
+	if (idx & 0x4)
+		return idx;
+	return GET_SWZ(swz, idx);
+}
+
+static inline GLuint combine_swizzles4(GLuint src, GLuint swz_x, GLuint swz_y, GLuint swz_z, GLuint swz_w)
+{
+	GLuint ret = 0;
+
+	ret |= get_swz(src, swz_x);
+	ret |= get_swz(src, swz_y) << 3;
+	ret |= get_swz(src, swz_z) << 6;
+	ret |= get_swz(src, swz_w) << 9;
+
+	return ret;
+}
+
+static inline GLuint combine_swizzles(GLuint src, GLuint swz)
+{
+	GLuint ret = 0;
+
+	ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_X));
+	ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_Y)) << 3;
+	ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_Z)) << 6;
+	ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_W)) << 9;
+
+	return ret;
+}
+
+
 /**
  * Transformation context that is passed to local transformations.
  *
-- 
cgit v1.2.3


From 582bd3466514b9fe24f18d99af2945f02709aacd Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sun, 12 Jul 2009 15:09:15 +0200
Subject: r300: always assume all components are read by fragment program
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Components of input attributes that are used by fragment program aren't part of vertex program key, and that may lead to situations when vertex program writes only TEX1.xy and fragment program reads TEX1.xyz, resulting in rendering errors.

Reported-by: Nicolai Hähnle <nhaehnle@gmail.com>
---
 src/mesa/drivers/dri/r300/r300_vertprog.c | 54 +++----------------------------
 1 file changed, 5 insertions(+), 49 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index c0b116156a..de32013032 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -1494,46 +1494,6 @@ static void addArtificialOutputs(GLcontext *ctx, struct gl_program *prog)
 
 #undef ADD_OUTPUT
 
-static GLuint getUsedComponents(const GLuint swizzle)
-{
-	GLuint ret;
-
-	ret = 0;
-
-	/* need to mask out ZERO, ONE and NIL swizzles */
-	if (GET_SWZ(swizzle, SWIZZLE_X) <= SWIZZLE_W)
-		ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_X));
-	if (GET_SWZ(swizzle, SWIZZLE_Y) <= SWIZZLE_W)
-		ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_Y));
-	if (GET_SWZ(swizzle, SWIZZLE_Z) <= SWIZZLE_W)
-		ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_Z));
-	if (GET_SWZ(swizzle, SWIZZLE_W) <= SWIZZLE_W)
-		ret |= 1 << (GET_SWZ(swizzle, SWIZZLE_W));
-
-	return ret;
-}
-
-static GLuint trackUsedComponents(const struct gl_program *prog, const GLuint attrib)
-{
-	struct prog_instruction *inst;
-	int tmp, i;
-	GLuint ret;
-
-	inst = prog->Instructions;
-	ret = 0;
-	while (inst->Opcode != OPCODE_END) {
-		tmp = _mesa_num_inst_src_regs(inst->Opcode);
-		for (i = 0; i < tmp; ++i) {
-			if (inst->SrcReg[i].File == PROGRAM_INPUT && inst->SrcReg[i].Index == attrib) {
-				ret |= getUsedComponents(inst->SrcReg[i].Swizzle);
-			}
-		}
-		++inst;
-	}
-
-	return ret;
-}
-
 static void nqssadceInit(struct nqssadce_state* s)
 {
 	r300ContextPtr r300 = R300_CONTEXT(s->Ctx);
@@ -1541,18 +1501,14 @@ static void nqssadceInit(struct nqssadce_state* s)
 
 	fp_reads = r300->selected_fp->Base->InputsRead;
 	{
-		GLuint tmp;
-
 		if (fp_reads & FRAG_BIT_COL0) {
-				tmp = trackUsedComponents(r300->selected_fp->Base, FRAG_ATTRIB_COL0);
-				s->Outputs[VERT_RESULT_COL0].Sourced = tmp;
-				s->Outputs[VERT_RESULT_BFC0].Sourced = tmp;
+				s->Outputs[VERT_RESULT_COL0].Sourced = WRITEMASK_XYZW;
+				s->Outputs[VERT_RESULT_BFC0].Sourced = WRITEMASK_XYZW;
 		}
 
 		if (fp_reads & FRAG_BIT_COL1) {
-				tmp = trackUsedComponents(r300->selected_fp->Base, FRAG_ATTRIB_COL1);
-				s->Outputs[VERT_RESULT_COL1].Sourced = tmp;
-				s->Outputs[VERT_RESULT_BFC1].Sourced = tmp;
+				s->Outputs[VERT_RESULT_COL1].Sourced = WRITEMASK_XYZW;
+				s->Outputs[VERT_RESULT_BFC1].Sourced = WRITEMASK_XYZW;
 		}
 	}
 
@@ -1560,7 +1516,7 @@ static void nqssadceInit(struct nqssadce_state* s)
 		int i;
 		for (i = 0; i < 8; ++i) {
 			if (fp_reads & FRAG_BIT_TEX(i)) {
-				s->Outputs[VERT_RESULT_TEX0 + i].Sourced = trackUsedComponents(r300->selected_fp->Base, FRAG_ATTRIB_TEX0 + i);
+				s->Outputs[VERT_RESULT_TEX0 + i].Sourced = WRITEMASK_XYZW;
 			}
 		}
 	}
-- 
cgit v1.2.3


From 60e60bb3026a269fefe1cfd3312fdf3a7e4c595f Mon Sep 17 00:00:00 2001
From: Michel Dänzer <daenzer@vmware.com>
Date: Tue, 14 Jul 2009 08:00:49 +0200
Subject: radeon: Invert front face winding when rendering to FBO.

Fixes fgl_glxgears and progs/demos/fbotexture after pressing 'c'.

Tested with r300, radeon and r200 compile tested only.
---
 src/mesa/drivers/dri/r200/r200_state.c     | 4 ++++
 src/mesa/drivers/dri/r300/r300_state.c     | 4 ++++
 src/mesa/drivers/dri/radeon/radeon_state.c | 4 ++++
 3 files changed, 12 insertions(+)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r200/r200_state.c b/src/mesa/drivers/dri/r200/r200_state.c
index c7c1a39316..5a6fd20d8c 100644
--- a/src/mesa/drivers/dri/r200/r200_state.c
+++ b/src/mesa/drivers/dri/r200/r200_state.c
@@ -574,6 +574,10 @@ static void r200FrontFace( GLcontext *ctx, GLenum mode )
    R200_STATECHANGE( rmesa, tcl );
    rmesa->hw.tcl.cmd[TCL_UCP_VERT_BLEND_CTL] &= ~R200_CULL_FRONT_IS_CCW;
 
+   /* Winding is inverted when rendering to FBO */
+   if (ctx->DrawBuffer && ctx->DrawBuffer->Name)
+      mode = (mode == GL_CW) ? GL_CCW : GL_CW;
+
    switch ( mode ) {
    case GL_CW:
       rmesa->hw.set.cmd[SET_SE_CNTL] |= R200_FFACE_CULL_CW;
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index 4d504d14e3..12fbf281d9 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -434,6 +434,10 @@ static void r300UpdateCulling(GLcontext * ctx)
 		break;
 	}
 
+	/* Winding is inverted when rendering to FBO */
+	if (ctx->DrawBuffer && ctx->DrawBuffer->Name)
+		val ^= R300_FRONT_FACE_CW;
+
 	R300_STATECHANGE(r300, cul);
 	r300->hw.cul.cmd[R300_CUL_CULL] = val;
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c
index 78f7e9d633..0d1728b747 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state.c
@@ -458,6 +458,10 @@ static void radeonFrontFace( GLcontext *ctx, GLenum mode )
    RADEON_STATECHANGE( rmesa, tcl );
    rmesa->hw.tcl.cmd[TCL_UCP_VERT_BLEND_CTL] &= ~RADEON_CULL_FRONT_IS_CCW;
 
+   /* Winding is inverted when rendering to FBO */
+   if (ctx->DrawBuffer && ctx->DrawBuffer->Name)
+      mode = (mode == GL_CW) ? GL_CCW : GL_CW;
+
    switch ( mode ) {
    case GL_CW:
       rmesa->hw.set.cmd[SET_SE_CNTL] |= RADEON_FFACE_CULL_CW;
-- 
cgit v1.2.3


From 33f56b4612e506999a2be8391ba82c0174afa1b3 Mon Sep 17 00:00:00 2001
From: Michel Dänzer <daenzer@vmware.com>
Date: Tue, 14 Jul 2009 08:25:27 +0200
Subject: radeon: Differentiate 16 bpp destination formats.

Fixes those formats in fbo_firecube.

Only tested with r300, radeon and r200 compile tested only.
---
 src/mesa/drivers/dri/r200/r200_state_init.c     | 11 ++++++++++-
 src/mesa/drivers/dri/r300/r300_cmdbuf.c         | 11 ++++++++++-
 src/mesa/drivers/dri/radeon/radeon_state_init.c | 11 ++++++++++-
 3 files changed, 30 insertions(+), 3 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r200/r200_state_init.c b/src/mesa/drivers/dri/r200/r200_state_init.c
index c7df4b2587..bc871d9904 100644
--- a/src/mesa/drivers/dri/r200/r200_state_init.c
+++ b/src/mesa/drivers/dri/r200/r200_state_init.c
@@ -484,8 +484,17 @@ static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
    atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10);
    if (rrb->cpp == 4)
 	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888;
-   else
+   else switch (rrb->base._ActualFormat) {
+   case GL_RGB5:
 	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565;
+	break;
+   case GL_RGBA4:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB4444;
+	break;
+   case GL_RGB5_A1:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB1555;
+	break;
+   }
 
    cbpitch = (rrb->pitch / rrb->cpp);
    if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
index 90ca45b3e7..248297897d 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
@@ -269,8 +269,17 @@ static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
 	cbpitch = (rrb->pitch / rrb->cpp);
 	if (rrb->cpp == 4)
 		cbpitch |= R300_COLOR_FORMAT_ARGB8888;
-	else
+	else switch (rrb->base._ActualFormat) {
+	case GL_RGB5:
 		cbpitch |= R300_COLOR_FORMAT_RGB565;
+		break;
+	case GL_RGBA4:
+		cbpitch |= R300_COLOR_FORMAT_ARGB4444;
+		break;
+	case GL_RGB5_A1:
+		cbpitch |= R300_COLOR_FORMAT_ARGB1555;
+		break;
+	}
 
 	if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
 		cbpitch |= R300_COLOR_TILE_ENABLE;
diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c
index a55cd6dfcb..aaa82b1d6a 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state_init.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c
@@ -390,8 +390,17 @@ static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
    atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10);
    if (rrb->cpp == 4)
 	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888;
-   else
+   else switch (rrb->base._ActualFormat) {
+   case GL_RGB5:
 	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565;
+	break;
+   case GL_RGBA4:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB4444;
+	break;
+   case GL_RGB5_A1:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB1555;
+	break;
+   }
 
    cbpitch = (rrb->pitch / rrb->cpp);
    if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
-- 
cgit v1.2.3


From f6f0e117a45a64464e49290ebc9f75b9a976070a Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Wed, 15 Jul 2009 09:35:09 +1000
Subject: intel/radeon: add common metaops code.

Move all the metaops to a dri_metaops file and port radeon/intel
to use the new common meta ops code.
---
 src/mesa/drivers/dri/Makefile.template             |   3 +-
 src/mesa/drivers/dri/common/dri_metaops.c          | 541 +++++++++++++++++++++
 src/mesa/drivers/dri/common/dri_metaops.h          |  99 ++++
 src/mesa/drivers/dri/intel/intel_clear.c           | 247 +---------
 src/mesa/drivers/dri/intel/intel_context.c         |   6 +-
 src/mesa/drivers/dri/intel/intel_context.h         |  33 +-
 src/mesa/drivers/dri/intel/intel_generatemipmap.c  |  18 +-
 src/mesa/drivers/dri/intel/intel_pixel.c           | 251 ----------
 src/mesa/drivers/dri/intel/intel_pixel.h           |  15 -
 src/mesa/drivers/dri/intel/intel_pixel_bitmap.c    |  16 +-
 src/mesa/drivers/dri/intel/intel_pixel_draw.c      |  16 +-
 src/mesa/drivers/dri/r200/r200_ioctl.c             |   7 +-
 src/mesa/drivers/dri/r300/r300_ioctl.c             |   7 +-
 src/mesa/drivers/dri/radeon/radeon_common.c        | 252 +---------
 src/mesa/drivers/dri/radeon/radeon_common.h        |  13 +-
 .../drivers/dri/radeon/radeon_common_context.c     |   3 +-
 .../drivers/dri/radeon/radeon_common_context.h     |  23 +-
 src/mesa/drivers/dri/radeon/radeon_ioctl.c         |   5 -
 18 files changed, 684 insertions(+), 871 deletions(-)
 create mode 100644 src/mesa/drivers/dri/common/dri_metaops.c
 create mode 100644 src/mesa/drivers/dri/common/dri_metaops.h

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/Makefile.template b/src/mesa/drivers/dri/Makefile.template
index bd38e3be47..18dbeba24a 100644
--- a/src/mesa/drivers/dri/Makefile.template
+++ b/src/mesa/drivers/dri/Makefile.template
@@ -11,7 +11,8 @@ COMMON_GALLIUM_SOURCES = \
 COMMON_SOURCES = $(COMMON_GALLIUM_SOURCES) \
         ../../common/driverfuncs.c \
         ../common/texmem.c \
-        ../common/drirenderbuffer.c
+        ../common/drirenderbuffer.c \
+	../common/dri_metaops.c
 
 ifeq ($(WINDOW_SYSTEM),dri)
 WINOBJ=
diff --git a/src/mesa/drivers/dri/common/dri_metaops.c b/src/mesa/drivers/dri/common/dri_metaops.c
new file mode 100644
index 0000000000..fe183c2e87
--- /dev/null
+++ b/src/mesa/drivers/dri/common/dri_metaops.c
@@ -0,0 +1,541 @@
+/**************************************************************************
+ *
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2009 Intel Corporation.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "main/arrayobj.h"
+#include "main/attrib.h"
+#include "main/blend.h"
+#include "main/bufferobj.h"
+#include "main/buffers.h"
+#include "main/depth.h"
+#include "main/enable.h"
+#include "main/matrix.h"
+#include "main/macros.h"
+#include "main/polygon.h"
+#include "main/shaders.h"
+#include "main/stencil.h"
+#include "main/texstate.h"
+#include "main/varray.h"
+#include "main/viewport.h"
+#include "shader/arbprogram.h"
+#include "shader/program.h"
+#include "dri_metaops.h"
+
+void
+meta_set_passthrough_transform(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+
+   meta->saved_vp_x = ctx->Viewport.X;
+   meta->saved_vp_y = ctx->Viewport.Y;
+   meta->saved_vp_width = ctx->Viewport.Width;
+   meta->saved_vp_height = ctx->Viewport.Height;
+   meta->saved_matrix_mode = ctx->Transform.MatrixMode;
+
+   meta->internal_viewport_call = GL_TRUE;
+   _mesa_Viewport(0, 0, ctx->DrawBuffer->Width, ctx->DrawBuffer->Height);
+   meta->internal_viewport_call = GL_FALSE;
+
+   _mesa_MatrixMode(GL_PROJECTION);
+   _mesa_PushMatrix();
+   _mesa_LoadIdentity();
+   _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1);
+
+   _mesa_MatrixMode(GL_MODELVIEW);
+   _mesa_PushMatrix();
+   _mesa_LoadIdentity();
+}
+
+void
+meta_restore_transform(struct dri_metaops *meta)
+{
+   _mesa_MatrixMode(GL_PROJECTION);
+   _mesa_PopMatrix();
+   _mesa_MatrixMode(GL_MODELVIEW);
+   _mesa_PopMatrix();
+
+   _mesa_MatrixMode(meta->saved_matrix_mode);
+
+   meta->internal_viewport_call = GL_TRUE;
+   _mesa_Viewport(meta->saved_vp_x, meta->saved_vp_y,
+		  meta->saved_vp_width, meta->saved_vp_height);
+   meta->internal_viewport_call = GL_FALSE;
+}
+
+
+/**
+ * Set up a vertex program to pass through the position and first texcoord
+ * for pixel path.
+ */
+void
+meta_set_passthrough_vertex_program(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+   static const char *vp =
+      "!!ARBvp1.0\n"
+      "TEMP vertexClip;\n"
+      "DP4 vertexClip.x, state.matrix.mvp.row[0], vertex.position;\n"
+      "DP4 vertexClip.y, state.matrix.mvp.row[1], vertex.position;\n"
+      "DP4 vertexClip.z, state.matrix.mvp.row[2], vertex.position;\n"
+      "DP4 vertexClip.w, state.matrix.mvp.row[3], vertex.position;\n"
+      "MOV result.position, vertexClip;\n"
+      "MOV result.texcoord[0], vertex.texcoord[0];\n"
+      "MOV result.color, vertex.color;\n"
+      "END\n";
+
+   assert(meta->saved_vp == NULL);
+
+   _mesa_reference_vertprog(ctx, &meta->saved_vp,
+			    ctx->VertexProgram.Current);
+   if (meta->passthrough_vp == NULL) {
+      GLuint prog_name;
+      _mesa_GenPrograms(1, &prog_name);
+      _mesa_BindProgram(GL_VERTEX_PROGRAM_ARB, prog_name);
+      _mesa_ProgramStringARB(GL_VERTEX_PROGRAM_ARB,
+			     GL_PROGRAM_FORMAT_ASCII_ARB,
+			     strlen(vp), (const GLubyte *)vp);
+      _mesa_reference_vertprog(ctx, &meta->passthrough_vp,
+			       ctx->VertexProgram.Current);
+      _mesa_DeletePrograms(1, &prog_name);
+   }
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
+			    meta->passthrough_vp);
+   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
+			   &meta->passthrough_vp->Base);
+
+   meta->saved_vp_enable = ctx->VertexProgram.Enabled;
+   _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
+}
+
+/**
+ * Restores the previous vertex program after
+ * meta_set_passthrough_vertex_program()
+ */
+void
+meta_restore_vertex_program(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
+			    meta->saved_vp);
+   _mesa_reference_vertprog(ctx, &meta->saved_vp, NULL);
+   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
+			   &ctx->VertexProgram.Current->Base);
+
+   if (!meta->saved_vp_enable)
+      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
+}
+
+/**
+ * Binds the given program string to GL_FRAGMENT_PROGRAM_ARB, caching the
+ * program object.
+ */
+void
+meta_set_fragment_program(struct dri_metaops *meta,
+			  struct gl_fragment_program **prog,
+			  const char *prog_string)
+{
+   GLcontext *ctx = meta->ctx;
+   assert(meta->saved_fp == NULL);
+
+   _mesa_reference_fragprog(ctx, &meta->saved_fp,
+			    ctx->FragmentProgram.Current);
+   if (*prog == NULL) {
+      GLuint prog_name;
+      _mesa_GenPrograms(1, &prog_name);
+      _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, prog_name);
+      _mesa_ProgramStringARB(GL_FRAGMENT_PROGRAM_ARB,
+			     GL_PROGRAM_FORMAT_ASCII_ARB,
+			     strlen(prog_string), (const GLubyte *)prog_string);
+      _mesa_reference_fragprog(ctx, prog, ctx->FragmentProgram.Current);
+      /* Note that DeletePrograms unbinds the program on us */
+      _mesa_DeletePrograms(1, &prog_name);
+   }
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, *prog);
+   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, &((*prog)->Base));
+
+   meta->saved_fp_enable = ctx->FragmentProgram.Enabled;
+   _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
+}
+
+/**
+ * Restores the previous fragment program after
+ * meta_set_fragment_program()
+ */
+void
+meta_restore_fragment_program(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current,
+			    meta->saved_fp);
+   _mesa_reference_fragprog(ctx, &meta->saved_fp, NULL);
+   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB,
+			   &ctx->FragmentProgram.Current->Base);
+
+   if (!meta->saved_fp_enable)
+      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
+}
+
+static const float default_texcoords[4][2] = { { 0.0, 0.0 },
+					       { 1.0, 0.0 },
+					       { 1.0, 1.0 },
+					       { 0.0, 1.0 } };
+
+void
+meta_set_default_texrect(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+   struct gl_client_array *old_texcoord_array;
+
+   meta->saved_active_texture = ctx->Texture.CurrentUnit;
+   if (meta->saved_array_vbo == NULL) {
+      _mesa_reference_buffer_object(ctx, &meta->saved_array_vbo,
+				    ctx->Array.ArrayBufferObj);
+   }
+
+   old_texcoord_array = &ctx->Array.ArrayObj->TexCoord[0];
+   meta->saved_texcoord_type = old_texcoord_array->Type;
+   meta->saved_texcoord_size = old_texcoord_array->Size;
+   meta->saved_texcoord_stride = old_texcoord_array->Stride;
+   meta->saved_texcoord_enable = old_texcoord_array->Enabled;
+   meta->saved_texcoord_ptr = old_texcoord_array->Ptr;
+   _mesa_reference_buffer_object(ctx, &meta->saved_texcoord_vbo,
+				 old_texcoord_array->BufferObj);
+
+   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
+
+   if (meta->texcoord_vbo == NULL) {
+      GLuint vbo_name;
+
+      _mesa_GenBuffersARB(1, &vbo_name);
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, vbo_name);
+      _mesa_BufferDataARB(GL_ARRAY_BUFFER_ARB, sizeof(default_texcoords),
+			  default_texcoords, GL_STATIC_DRAW_ARB);
+      _mesa_reference_buffer_object(ctx, &meta->texcoord_vbo,
+				    ctx->Array.ArrayBufferObj);
+   } else {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
+			  meta->texcoord_vbo->Name);
+   }
+   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), NULL);
+
+   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
+}
+
+void
+meta_restore_texcoords(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+
+   /* Restore the old TexCoordPointer */
+   if (meta->saved_texcoord_vbo) {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
+			  meta->saved_texcoord_vbo->Name);
+      _mesa_reference_buffer_object(ctx, &meta->saved_texcoord_vbo, NULL);
+   } else {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
+   }
+
+   _mesa_TexCoordPointer(meta->saved_texcoord_size,
+			 meta->saved_texcoord_type,
+			 meta->saved_texcoord_stride,
+			 meta->saved_texcoord_ptr);
+   if (!meta->saved_texcoord_enable)
+      _mesa_Disable(GL_TEXTURE_COORD_ARRAY);
+
+   _mesa_ClientActiveTextureARB(GL_TEXTURE0 +
+				meta->saved_active_texture);
+
+   if (meta->saved_array_vbo) {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
+			  meta->saved_array_vbo->Name);
+      _mesa_reference_buffer_object(ctx, &meta->saved_array_vbo, NULL);
+   } else {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
+   }
+}
+
+
+/**
+ * Perform glClear where mask contains only color, depth, and/or stencil.
+ *
+ * The implementation is based on calling into Mesa to set GL state and
+ * performing normal triangle rendering.  The intent of this path is to
+ * have as generic a path as possible, so that any driver could make use of
+ * it.
+ */
+
+/**
+ * Per-context one-time init of things for intl_clear_tris().
+ * Basically set up a private array object for vertex/color arrays.
+ */
+static void
+meta_init_clear(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+   struct gl_array_object *arraySave = NULL;
+   const GLuint arrayBuffer = ctx->Array.ArrayBufferObj->Name;
+   const GLuint elementBuffer = ctx->Array.ElementArrayBufferObj->Name;
+
+   /* create new array object */
+   meta->clear.arrayObj = _mesa_new_array_object(ctx, ~0);
+
+   /* save current array object, bind new one */
+   _mesa_reference_array_object(ctx, &arraySave, ctx->Array.ArrayObj);
+   ctx->NewState |= _NEW_ARRAY;
+   ctx->Array.NewState |= _NEW_ARRAY_ALL;
+   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, meta->clear.arrayObj);
+
+   /* one-time setup of vertex arrays (pos, color) */
+   _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
+   _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0);
+   _mesa_ColorPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), meta->clear.color);
+   _mesa_VertexPointer(3, GL_FLOAT, 3 * sizeof(GLfloat), meta->clear.vertices);
+   _mesa_Enable(GL_COLOR_ARRAY);
+   _mesa_Enable(GL_VERTEX_ARRAY);
+
+   /* restore original array object */
+   ctx->NewState |= _NEW_ARRAY;
+   ctx->Array.NewState |= _NEW_ARRAY_ALL;
+   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, arraySave);
+   _mesa_reference_array_object(ctx, &arraySave, NULL);
+
+   /* restore original buffer objects */
+   _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, arrayBuffer);
+   _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, elementBuffer);
+}
+
+
+
+/**
+ * Perform glClear where mask contains only color, depth, and/or stencil.
+ *
+ * The implementation is based on calling into Mesa to set GL state and
+ * performing normal triangle rendering.  The intent of this path is to
+ * have as generic a path as possible, so that any driver could make use of
+ * it.
+ */
+void
+meta_clear_tris(struct dri_metaops *meta, GLbitfield mask)
+{
+   GLcontext *ctx = meta->ctx;
+   GLfloat dst_z;
+   struct gl_framebuffer *fb = ctx->DrawBuffer;
+   int i;
+   GLboolean saved_fp_enable = GL_FALSE, saved_vp_enable = GL_FALSE;
+   GLuint saved_shader_program = 0;
+   unsigned int saved_active_texture;
+   struct gl_array_object *arraySave = NULL;
+
+   if (!meta->clear.arrayObj)
+      meta_init_clear(meta);
+
+   assert((mask & ~(TRI_CLEAR_COLOR_BITS | BUFFER_BIT_DEPTH |
+		    BUFFER_BIT_STENCIL)) == 0);
+
+   _mesa_PushAttrib(GL_COLOR_BUFFER_BIT |
+		    GL_DEPTH_BUFFER_BIT |
+		    GL_ENABLE_BIT |
+		    GL_POLYGON_BIT |
+		    GL_STENCIL_BUFFER_BIT |
+		    GL_TRANSFORM_BIT |
+		    GL_CURRENT_BIT |
+		    GL_VIEWPORT_BIT);
+   saved_active_texture = ctx->Texture.CurrentUnit;
+
+   /* Disable existing GL state we don't want to apply to a clear. */
+   _mesa_Disable(GL_ALPHA_TEST);
+   _mesa_Disable(GL_BLEND);
+   _mesa_Disable(GL_CULL_FACE);
+   _mesa_Disable(GL_FOG);
+   _mesa_Disable(GL_POLYGON_SMOOTH);
+   _mesa_Disable(GL_POLYGON_STIPPLE);
+   _mesa_Disable(GL_POLYGON_OFFSET_FILL);
+   _mesa_Disable(GL_LIGHTING);
+   _mesa_Disable(GL_CLIP_PLANE0);
+   _mesa_Disable(GL_CLIP_PLANE1);
+   _mesa_Disable(GL_CLIP_PLANE2);
+   _mesa_Disable(GL_CLIP_PLANE3);
+   _mesa_Disable(GL_CLIP_PLANE4);
+   _mesa_Disable(GL_CLIP_PLANE5);
+   _mesa_PolygonMode(GL_FRONT_AND_BACK, GL_FILL);
+   if (ctx->Extensions.ARB_fragment_program && ctx->FragmentProgram.Enabled) {
+      saved_fp_enable = GL_TRUE;
+      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
+   }
+   if (ctx->Extensions.ARB_vertex_program && ctx->VertexProgram.Enabled) {
+      saved_vp_enable = GL_TRUE;
+      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
+   }
+   if (ctx->Extensions.ARB_shader_objects && ctx->Shader.CurrentProgram) {
+      saved_shader_program = ctx->Shader.CurrentProgram->Name;
+      _mesa_UseProgramObjectARB(0);
+   }
+
+   if (ctx->Texture._EnabledUnits != 0) {
+      int i;
+
+      for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
+	 _mesa_ActiveTextureARB(GL_TEXTURE0 + i);
+	 _mesa_Disable(GL_TEXTURE_1D);
+	 _mesa_Disable(GL_TEXTURE_2D);
+	 _mesa_Disable(GL_TEXTURE_3D);
+	 if (ctx->Extensions.ARB_texture_cube_map)
+	    _mesa_Disable(GL_TEXTURE_CUBE_MAP_ARB);
+	 if (ctx->Extensions.NV_texture_rectangle)
+	    _mesa_Disable(GL_TEXTURE_RECTANGLE_NV);
+	 if (ctx->Extensions.MESA_texture_array) {
+	    _mesa_Disable(GL_TEXTURE_1D_ARRAY_EXT);
+	    _mesa_Disable(GL_TEXTURE_2D_ARRAY_EXT);
+	 }
+      }
+   }
+
+   /* save current array object, bind our private one */
+   _mesa_reference_array_object(ctx, &arraySave, ctx->Array.ArrayObj);
+   ctx->NewState |= _NEW_ARRAY;
+   ctx->Array.NewState |= _NEW_ARRAY_ALL;
+   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, meta->clear.arrayObj);
+
+   meta_set_passthrough_transform(meta);
+
+   for (i = 0; i < 4; i++) {
+      COPY_4FV(meta->clear.color[i], ctx->Color.ClearColor);
+   }
+
+   /* convert clear Z from [0,1] to NDC coord in [-1,1] */
+   dst_z = -1.0 + 2.0 * ctx->Depth.Clear;
+
+   /* The ClearDepth value is unaffected by DepthRange, so do a default
+    * mapping.
+    */
+   _mesa_DepthRange(0.0, 1.0);
+
+   /* Prepare the vertices, which are the same regardless of which buffer we're
+    * drawing to.
+    */
+   meta->clear.vertices[0][0] = fb->_Xmin;
+   meta->clear.vertices[0][1] = fb->_Ymin;
+   meta->clear.vertices[0][2] = dst_z;
+   meta->clear.vertices[1][0] = fb->_Xmax;
+   meta->clear.vertices[1][1] = fb->_Ymin;
+   meta->clear.vertices[1][2] = dst_z;
+   meta->clear.vertices[2][0] = fb->_Xmax;
+   meta->clear.vertices[2][1] = fb->_Ymax;
+   meta->clear.vertices[2][2] = dst_z;
+   meta->clear.vertices[3][0] = fb->_Xmin;
+   meta->clear.vertices[3][1] = fb->_Ymax;
+   meta->clear.vertices[3][2] = dst_z;
+
+   while (mask != 0) {
+      GLuint this_mask = 0;
+      GLuint color_bit;
+
+      color_bit = _mesa_ffs(mask & TRI_CLEAR_COLOR_BITS);
+      if (color_bit != 0)
+	 this_mask |= (1 << (color_bit - 1));
+
+      /* Clear depth/stencil in the same pass as color. */
+      this_mask |= (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL));
+
+      /* Select the current color buffer and use the color write mask if
+       * we have one, otherwise don't write any color channels.
+       */
+      if (this_mask & BUFFER_BIT_FRONT_LEFT)
+	 _mesa_DrawBuffer(GL_FRONT_LEFT);
+      else if (this_mask & BUFFER_BIT_BACK_LEFT)
+	 _mesa_DrawBuffer(GL_BACK_LEFT);
+      else if (color_bit != 0)
+	 _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0 +
+			  (color_bit - BUFFER_COLOR0 - 1));
+      else
+	 _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
+
+      /* Control writing of the depth clear value to depth. */
+      if (this_mask & BUFFER_BIT_DEPTH) {
+	 _mesa_DepthFunc(GL_ALWAYS);
+	 _mesa_Enable(GL_DEPTH_TEST);
+      } else {
+	 _mesa_Disable(GL_DEPTH_TEST);
+	 _mesa_DepthMask(GL_FALSE);
+      }
+
+      /* Control writing of the stencil clear value to stencil. */
+      if (this_mask & BUFFER_BIT_STENCIL) {
+	 _mesa_Enable(GL_STENCIL_TEST);
+	 _mesa_StencilOpSeparate(GL_FRONT_AND_BACK,
+				 GL_REPLACE, GL_REPLACE, GL_REPLACE);
+	 _mesa_StencilFuncSeparate(GL_FRONT_AND_BACK, GL_ALWAYS,
+				   ctx->Stencil.Clear & 0x7fffffff,
+				   ctx->Stencil.WriteMask[0]);
+      } else {
+	 _mesa_Disable(GL_STENCIL_TEST);
+      }
+
+      _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
+
+      mask &= ~this_mask;
+   }
+
+   meta_restore_transform(meta);
+
+   _mesa_ActiveTextureARB(GL_TEXTURE0 + saved_active_texture);
+   if (saved_fp_enable)
+      _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
+   if (saved_vp_enable)
+      _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
+
+   if (saved_shader_program)
+      _mesa_UseProgramObjectARB(saved_shader_program);
+
+   _mesa_PopAttrib();
+
+   /* restore current array object */
+   ctx->NewState |= _NEW_ARRAY;
+   ctx->Array.NewState |= _NEW_ARRAY_ALL;
+   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, arraySave);
+   _mesa_reference_array_object(ctx, &arraySave, NULL);
+}
+
+void meta_init_metaops(GLcontext *ctx, struct dri_metaops *meta)
+{
+   meta->ctx = ctx;
+}
+
+void meta_destroy_metaops(struct dri_metaops *meta)
+{
+  if (meta->clear.arrayObj)
+    _mesa_delete_array_object(meta->ctx, meta->clear.arrayObj);
+
+}
diff --git a/src/mesa/drivers/dri/common/dri_metaops.h b/src/mesa/drivers/dri/common/dri_metaops.h
new file mode 100644
index 0000000000..bb4079d535
--- /dev/null
+++ b/src/mesa/drivers/dri/common/dri_metaops.h
@@ -0,0 +1,99 @@
+/**************************************************************************
+ *
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2009 Intel Corporation.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+#ifndef DRI_METAOPS_H
+#define DRI_METAOPS_H
+
+#define TRI_CLEAR_COLOR_BITS (BUFFER_BIT_BACK_LEFT |			\
+			      BUFFER_BIT_FRONT_LEFT |			\
+			      BUFFER_BIT_COLOR0 |			\
+			      BUFFER_BIT_COLOR1 |			\
+			      BUFFER_BIT_COLOR2 |			\
+			      BUFFER_BIT_COLOR3 |			\
+			      BUFFER_BIT_COLOR4 |			\
+			      BUFFER_BIT_COLOR5 |			\
+			      BUFFER_BIT_COLOR6 |			\
+			      BUFFER_BIT_COLOR7)
+
+struct dri_meta_clear {
+    struct gl_array_object *arrayObj;
+    GLfloat vertices[4][3];
+    GLfloat color[4][4];
+};
+
+struct dri_metaops {
+    GLcontext *ctx;
+    GLboolean internal_viewport_call;
+    struct gl_fragment_program *bitmap_fp;
+    struct gl_vertex_program *passthrough_vp;
+    struct gl_buffer_object *texcoord_vbo;
+    
+    struct gl_fragment_program *saved_fp;
+    GLboolean saved_fp_enable;
+    struct gl_vertex_program *saved_vp;
+    GLboolean saved_vp_enable;
+
+    struct gl_fragment_program *tex2d_fp;
+    
+    GLboolean saved_texcoord_enable;
+    struct gl_buffer_object *saved_array_vbo, *saved_texcoord_vbo;
+    GLenum saved_texcoord_type;
+    GLsizei saved_texcoord_size, saved_texcoord_stride;
+    const void *saved_texcoord_ptr;
+    int saved_active_texture;
+
+    GLint saved_vp_x, saved_vp_y;
+    GLsizei saved_vp_width, saved_vp_height;
+    GLenum saved_matrix_mode;
+    
+    struct dri_meta_clear clear;
+};
+
+
+void meta_set_passthrough_transform(struct dri_metaops *meta);
+
+void meta_restore_transform(struct dri_metaops *meta);
+
+void meta_set_passthrough_vertex_program(struct dri_metaops *meta);
+
+void meta_restore_vertex_program(struct dri_metaops *meta);
+
+void meta_set_fragment_program(struct dri_metaops *meta,
+			  struct gl_fragment_program **prog,
+			  const char *prog_string);
+
+void meta_restore_fragment_program(struct dri_metaops *meta);
+
+void meta_set_default_texrect(struct dri_metaops *meta);
+
+void meta_restore_texcoords(struct dri_metaops *meta);
+void meta_clear_tris(struct dri_metaops *meta, GLbitfield mask);
+
+void meta_init_metaops(GLcontext *ctx, struct dri_metaops *meta);
+void meta_destroy_metaops(struct dri_metaops *meta);
+#endif
+
diff --git a/src/mesa/drivers/dri/intel/intel_clear.c b/src/mesa/drivers/dri/intel/intel_clear.c
index 13b433dd17..cfddabd318 100644
--- a/src/mesa/drivers/dri/intel/intel_clear.c
+++ b/src/mesa/drivers/dri/intel/intel_clear.c
@@ -57,250 +57,6 @@
 
 #define FILE_DEBUG_FLAG DEBUG_BLIT
 
-#define TRI_CLEAR_COLOR_BITS (BUFFER_BIT_BACK_LEFT |			\
-			      BUFFER_BIT_FRONT_LEFT |			\
-			      BUFFER_BIT_COLOR0 |			\
-			      BUFFER_BIT_COLOR1 |			\
-			      BUFFER_BIT_COLOR2 |			\
-			      BUFFER_BIT_COLOR3 |			\
-			      BUFFER_BIT_COLOR4 |			\
-			      BUFFER_BIT_COLOR5 |			\
-			      BUFFER_BIT_COLOR6 |			\
-			      BUFFER_BIT_COLOR7)
-
-
-/**
- * Per-context one-time init of things for intl_clear_tris().
- * Basically set up a private array object for vertex/color arrays.
- */
-static void
-init_clear(GLcontext *ctx)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct gl_array_object *arraySave = NULL;
-   const GLuint arrayBuffer = ctx->Array.ArrayBufferObj->Name;
-   const GLuint elementBuffer = ctx->Array.ElementArrayBufferObj->Name;
-
-   /* create new array object */
-   intel->clear.arrayObj = _mesa_new_array_object(ctx, ~0);
-
-   /* save current array object, bind new one */
-   _mesa_reference_array_object(ctx, &arraySave, ctx->Array.ArrayObj);
-   ctx->NewState |= _NEW_ARRAY;
-   ctx->Array.NewState |= _NEW_ARRAY_ALL;
-   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, intel->clear.arrayObj);
-
-   /* one-time setup of vertex arrays (pos, color) */
-   _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
-   _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0);
-   _mesa_ColorPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), intel->clear.color);
-   _mesa_VertexPointer(3, GL_FLOAT, 3 * sizeof(GLfloat), intel->clear.vertices);
-   _mesa_Enable(GL_COLOR_ARRAY);
-   _mesa_Enable(GL_VERTEX_ARRAY);
-
-   /* restore original array object */
-   ctx->NewState |= _NEW_ARRAY;
-   ctx->Array.NewState |= _NEW_ARRAY_ALL;
-   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, arraySave);
-   _mesa_reference_array_object(ctx, &arraySave, NULL);
-
-   /* restore original buffer objects */
-   _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, arrayBuffer);
-   _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, elementBuffer);
-}
-
-
-
-/**
- * Perform glClear where mask contains only color, depth, and/or stencil.
- *
- * The implementation is based on calling into Mesa to set GL state and
- * performing normal triangle rendering.  The intent of this path is to
- * have as generic a path as possible, so that any driver could make use of
- * it.
- */
-void
-intel_clear_tris(GLcontext *ctx, GLbitfield mask)
-{
-   struct intel_context *intel = intel_context(ctx);
-   GLfloat dst_z;
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   int i;
-   GLboolean saved_fp_enable = GL_FALSE, saved_vp_enable = GL_FALSE;
-   GLuint saved_shader_program = 0;
-   unsigned int saved_active_texture;
-   struct gl_array_object *arraySave = NULL;
-
-   if (!intel->clear.arrayObj)
-      init_clear(ctx);
-
-   assert((mask & ~(TRI_CLEAR_COLOR_BITS | BUFFER_BIT_DEPTH |
-		    BUFFER_BIT_STENCIL)) == 0);
-
-   _mesa_PushAttrib(GL_COLOR_BUFFER_BIT |
-		    GL_DEPTH_BUFFER_BIT |
-		    GL_ENABLE_BIT |
-		    GL_POLYGON_BIT |
-		    GL_STENCIL_BUFFER_BIT |
-		    GL_TRANSFORM_BIT |
-		    GL_CURRENT_BIT |
-		    GL_VIEWPORT_BIT);
-   saved_active_texture = ctx->Texture.CurrentUnit;
-
-   /* Disable existing GL state we don't want to apply to a clear. */
-   _mesa_Disable(GL_ALPHA_TEST);
-   _mesa_Disable(GL_BLEND);
-   _mesa_Disable(GL_CULL_FACE);
-   _mesa_Disable(GL_FOG);
-   _mesa_Disable(GL_POLYGON_SMOOTH);
-   _mesa_Disable(GL_POLYGON_STIPPLE);
-   _mesa_Disable(GL_POLYGON_OFFSET_FILL);
-   _mesa_Disable(GL_LIGHTING);
-   _mesa_Disable(GL_CLIP_PLANE0);
-   _mesa_Disable(GL_CLIP_PLANE1);
-   _mesa_Disable(GL_CLIP_PLANE2);
-   _mesa_Disable(GL_CLIP_PLANE3);
-   _mesa_Disable(GL_CLIP_PLANE4);
-   _mesa_Disable(GL_CLIP_PLANE5);
-   _mesa_PolygonMode(GL_FRONT_AND_BACK, GL_FILL);
-   if (ctx->Extensions.ARB_fragment_program && ctx->FragmentProgram.Enabled) {
-      saved_fp_enable = GL_TRUE;
-      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
-   }
-   if (ctx->Extensions.ARB_vertex_program && ctx->VertexProgram.Enabled) {
-      saved_vp_enable = GL_TRUE;
-      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
-   }
-   if (ctx->Extensions.ARB_shader_objects && ctx->Shader.CurrentProgram) {
-      saved_shader_program = ctx->Shader.CurrentProgram->Name;
-      _mesa_UseProgramObjectARB(0);
-   }
-
-   if (ctx->Texture._EnabledUnits != 0) {
-      int i;
-
-      for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-	 _mesa_ActiveTextureARB(GL_TEXTURE0 + i);
-	 _mesa_Disable(GL_TEXTURE_1D);
-	 _mesa_Disable(GL_TEXTURE_2D);
-	 _mesa_Disable(GL_TEXTURE_3D);
-	 if (ctx->Extensions.ARB_texture_cube_map)
-	    _mesa_Disable(GL_TEXTURE_CUBE_MAP_ARB);
-	 if (ctx->Extensions.NV_texture_rectangle)
-	    _mesa_Disable(GL_TEXTURE_RECTANGLE_NV);
-	 if (ctx->Extensions.MESA_texture_array) {
-	    _mesa_Disable(GL_TEXTURE_1D_ARRAY_EXT);
-	    _mesa_Disable(GL_TEXTURE_2D_ARRAY_EXT);
-	 }
-      }
-   }
-
-   /* save current array object, bind our private one */
-   _mesa_reference_array_object(ctx, &arraySave, ctx->Array.ArrayObj);
-   ctx->NewState |= _NEW_ARRAY;
-   ctx->Array.NewState |= _NEW_ARRAY_ALL;
-   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, intel->clear.arrayObj);
-
-   intel_meta_set_passthrough_transform(intel);
-
-   for (i = 0; i < 4; i++) {
-      COPY_4FV(intel->clear.color[i], ctx->Color.ClearColor);
-   }
-
-   /* convert clear Z from [0,1] to NDC coord in [-1,1] */
-   dst_z = -1.0 + 2.0 * ctx->Depth.Clear;
-
-   /* The ClearDepth value is unaffected by DepthRange, so do a default
-    * mapping.
-    */
-   _mesa_DepthRange(0.0, 1.0);
-
-   /* Prepare the vertices, which are the same regardless of which buffer we're
-    * drawing to.
-    */
-   intel->clear.vertices[0][0] = fb->_Xmin;
-   intel->clear.vertices[0][1] = fb->_Ymin;
-   intel->clear.vertices[0][2] = dst_z;
-   intel->clear.vertices[1][0] = fb->_Xmax;
-   intel->clear.vertices[1][1] = fb->_Ymin;
-   intel->clear.vertices[1][2] = dst_z;
-   intel->clear.vertices[2][0] = fb->_Xmax;
-   intel->clear.vertices[2][1] = fb->_Ymax;
-   intel->clear.vertices[2][2] = dst_z;
-   intel->clear.vertices[3][0] = fb->_Xmin;
-   intel->clear.vertices[3][1] = fb->_Ymax;
-   intel->clear.vertices[3][2] = dst_z;
-
-   while (mask != 0) {
-      GLuint this_mask = 0;
-      GLuint color_bit;
-
-      color_bit = _mesa_ffs(mask & TRI_CLEAR_COLOR_BITS);
-      if (color_bit != 0)
-	 this_mask |= (1 << (color_bit - 1));
-
-      /* Clear depth/stencil in the same pass as color. */
-      this_mask |= (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL));
-
-      /* Select the current color buffer and use the color write mask if
-       * we have one, otherwise don't write any color channels.
-       */
-      if (this_mask & BUFFER_BIT_FRONT_LEFT)
-	 _mesa_DrawBuffer(GL_FRONT_LEFT);
-      else if (this_mask & BUFFER_BIT_BACK_LEFT)
-	 _mesa_DrawBuffer(GL_BACK_LEFT);
-      else if (color_bit != 0)
-	 _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0 +
-			  (color_bit - BUFFER_COLOR0 - 1));
-      else
-	 _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
-
-      /* Control writing of the depth clear value to depth. */
-      if (this_mask & BUFFER_BIT_DEPTH) {
-	 _mesa_DepthFunc(GL_ALWAYS);
-	 _mesa_Enable(GL_DEPTH_TEST);
-      } else {
-	 _mesa_Disable(GL_DEPTH_TEST);
-	 _mesa_DepthMask(GL_FALSE);
-      }
-
-      /* Control writing of the stencil clear value to stencil. */
-      if (this_mask & BUFFER_BIT_STENCIL) {
-	 _mesa_Enable(GL_STENCIL_TEST);
-	 _mesa_StencilOpSeparate(GL_FRONT_AND_BACK,
-				 GL_REPLACE, GL_REPLACE, GL_REPLACE);
-	 _mesa_StencilFuncSeparate(GL_FRONT_AND_BACK, GL_ALWAYS,
-				   ctx->Stencil.Clear,
-				   ctx->Stencil.WriteMask[0]);
-      } else {
-	 _mesa_Disable(GL_STENCIL_TEST);
-      }
-
-      _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
-
-      mask &= ~this_mask;
-   }
-
-   intel_meta_restore_transform(intel);
-
-   _mesa_ActiveTextureARB(GL_TEXTURE0 + saved_active_texture);
-   if (saved_fp_enable)
-      _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
-   if (saved_vp_enable)
-      _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
-
-   if (saved_shader_program)
-      _mesa_UseProgramObjectARB(saved_shader_program);
-
-   _mesa_PopAttrib();
-
-   /* restore current array object */
-   ctx->NewState |= _NEW_ARRAY;
-   ctx->Array.NewState |= _NEW_ARRAY_ALL;
-   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, arraySave);
-   _mesa_reference_array_object(ctx, &arraySave, NULL);
-}
-
 static const char *buffer_names[] = {
    [BUFFER_FRONT_LEFT] = "front",
    [BUFFER_BACK_LEFT] = "back",
@@ -326,6 +82,7 @@ static const char *buffer_names[] = {
 static void
 intelClear(GLcontext *ctx, GLbitfield mask)
 {
+   struct intel_context *intel = intel_context(ctx);
    const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask);
    GLbitfield tri_mask = 0;
    GLbitfield blit_mask = 0;
@@ -425,7 +182,7 @@ intelClear(GLcontext *ctx, GLbitfield mask)
 	 }
 	 DBG("\n");
       }
-      intel_clear_tris(ctx, tri_mask);
+      meta_clear_tris(&intel->meta, tri_mask);
    }
 
    if (swrast_mask) {
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index edf422c1bd..4abb525f78 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -404,7 +404,7 @@ intel_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
     if (!driContext->driScreenPriv->dri2.enabled)
 	return;
 
-    if (!intel->internal_viewport_call && ctx->DrawBuffer->Name == 0) {
+    if (!intel->meta.internal_viewport_call && ctx->DrawBuffer->Name == 0) {
        /* If we're rendering to the fake front buffer, make sure all the pending
 	* drawing has landed on the real front buffer.  Otherwise when we
 	* eventually get to DRI2GetBuffersWithFormat the stale real front
@@ -672,6 +672,7 @@ intelInitContext(struct intel_context *intel,
     */
    _mesa_init_point(ctx);
 
+   meta_init_metaops(ctx, &intel->meta);
    ctx->Const.MaxColorAttachments = 4;  /* XXX FBO: review this */
    if (IS_965(intelScreen->deviceID)) {
       if (MAX_WIDTH > 8192)
@@ -794,8 +795,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
 
       INTEL_FIREVERTICES(intel);
 
-      if (intel->clear.arrayObj)
-         _mesa_delete_array_object(&intel->ctx, intel->clear.arrayObj);
+      meta_destroy_metaops(&intel->meta);
 
       intel->vtbl.destroy(intel);
 
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index 6761193f8c..08bea88c95 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -33,6 +33,7 @@
 #include "main/mtypes.h"
 #include "main/mm.h"
 #include "texmem.h"
+#include "dri_metaops.h"
 #include "drm.h"
 #include "intel_bufmgr.h"
 
@@ -157,29 +158,7 @@ struct intel_context
       void (*debug_batch)(struct intel_context *intel);
    } vtbl;
 
-   struct {
-      struct gl_fragment_program *bitmap_fp;
-      struct gl_vertex_program *passthrough_vp;
-      struct gl_buffer_object *texcoord_vbo;
-
-      struct gl_fragment_program *saved_fp;
-      GLboolean saved_fp_enable;
-      struct gl_vertex_program *saved_vp;
-      GLboolean saved_vp_enable;
-
-      struct gl_fragment_program *tex2d_fp;
-
-      GLboolean saved_texcoord_enable;
-      struct gl_buffer_object *saved_array_vbo, *saved_texcoord_vbo;
-      GLenum saved_texcoord_type;
-      GLsizei saved_texcoord_size, saved_texcoord_stride;
-      const void *saved_texcoord_ptr;
-      int saved_active_texture;
-
-      GLint saved_vp_x, saved_vp_y;
-      GLsizei saved_vp_width, saved_vp_height;
-      GLenum saved_matrix_mode;
-   } meta;
+   struct dri_metaops meta;
 
    GLint refcount;
    GLuint Fallback;
@@ -191,7 +170,6 @@ struct intel_context
    struct intel_region *front_region;
    struct intel_region *back_region;
    struct intel_region *depth_region;
-   GLboolean internal_viewport_call;
 
    /**
     * This value indicates that the kernel memory manager is being used
@@ -224,13 +202,6 @@ struct intel_context
    GLuint ClearColor565;
    GLuint ClearColor8888;
 
-   /* info for intel_clear_tris() */
-   struct
-   {
-      struct gl_array_object *arrayObj;
-      GLfloat vertices[4][3];
-      GLfloat color[4][4];
-   } clear;
 
    /* Offsets of fields within the current vertex:
     */
diff --git a/src/mesa/drivers/dri/intel/intel_generatemipmap.c b/src/mesa/drivers/dri/intel/intel_generatemipmap.c
index b00f8019dd..fe986092db 100644
--- a/src/mesa/drivers/dri/intel/intel_generatemipmap.c
+++ b/src/mesa/drivers/dri/intel/intel_generatemipmap.c
@@ -88,7 +88,7 @@ intel_generate_mipmap_level(GLcontext *ctx, GLuint tex_name,
    if (status != GL_FRAMEBUFFER_COMPLETE_EXT)
       return GL_FALSE;
 
-   intel_meta_set_passthrough_transform(intel);
+   meta_set_passthrough_transform(&intel->meta);
 
    /* XXX: Doing it right would involve setting up the transformation to do
     * 0-1 mapping or something, and not changing the vertex data.
@@ -104,12 +104,12 @@ intel_generate_mipmap_level(GLcontext *ctx, GLuint tex_name,
 
    _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices);
    _mesa_Enable(GL_VERTEX_ARRAY);
-   intel_meta_set_default_texrect(intel);
+   meta_set_default_texrect(&intel->meta);
 
    _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
 
-   intel_meta_restore_texcoords(intel);
-   intel_meta_restore_transform(intel);
+   meta_restore_texcoords(&intel->meta);
+   meta_restore_transform(&intel->meta);
 
    return GL_TRUE;
 }
@@ -153,9 +153,9 @@ intel_generate_mipmap_2d(GLcontext *ctx,
    _mesa_GenFramebuffersEXT(1, &fb_name);
    _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, fb_name);
 
-   intel_meta_set_fragment_program(intel, &intel->meta.tex2d_fp,
-				   intel_fp_tex2d);
-   intel_meta_set_passthrough_vertex_program(intel);
+   meta_set_fragment_program(&intel->meta, &intel->meta.tex2d_fp,
+			     intel_fp_tex2d);
+   meta_set_passthrough_vertex_program(&intel->meta);
 
    max_levels = _mesa_max_texture_levels(ctx, texObj->Target);
    start_level = texObj->BaseLevel;
@@ -202,8 +202,8 @@ intel_generate_mipmap_2d(GLcontext *ctx,
    success = GL_TRUE;
 
 fail:
-   intel_meta_restore_fragment_program(intel);
-   intel_meta_restore_vertex_program(intel);
+   meta_restore_fragment_program(&intel->meta);
+   meta_restore_vertex_program(&intel->meta);
 
    _mesa_DeleteFramebuffersEXT(1, &fb_name);
    _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.c b/src/mesa/drivers/dri/intel/intel_pixel.c
index da9ccb23f1..a300141655 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel.c
@@ -176,246 +176,6 @@ intel_check_blit_format(struct intel_region * region,
    return GL_FALSE;
 }
 
-void
-intel_meta_set_passthrough_transform(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   intel->meta.saved_vp_x = ctx->Viewport.X;
-   intel->meta.saved_vp_y = ctx->Viewport.Y;
-   intel->meta.saved_vp_width = ctx->Viewport.Width;
-   intel->meta.saved_vp_height = ctx->Viewport.Height;
-   intel->meta.saved_matrix_mode = ctx->Transform.MatrixMode;
-
-   intel->internal_viewport_call = GL_TRUE;
-   _mesa_Viewport(0, 0, ctx->DrawBuffer->Width, ctx->DrawBuffer->Height);
-   intel->internal_viewport_call = GL_FALSE;
-
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-   _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1);
-
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-}
-
-void
-intel_meta_restore_transform(struct intel_context *intel)
-{
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PopMatrix();
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PopMatrix();
-
-   _mesa_MatrixMode(intel->meta.saved_matrix_mode);
-
-   intel->internal_viewport_call = GL_TRUE;
-   _mesa_Viewport(intel->meta.saved_vp_x, intel->meta.saved_vp_y,
-		  intel->meta.saved_vp_width, intel->meta.saved_vp_height);
-   intel->internal_viewport_call = GL_FALSE;
-}
-
-/**
- * Set up a vertex program to pass through the position and first texcoord
- * for pixel path.
- */
-void
-intel_meta_set_passthrough_vertex_program(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-   static const char *vp =
-      "!!ARBvp1.0\n"
-      "TEMP vertexClip;\n"
-      "DP4 vertexClip.x, state.matrix.mvp.row[0], vertex.position;\n"
-      "DP4 vertexClip.y, state.matrix.mvp.row[1], vertex.position;\n"
-      "DP4 vertexClip.z, state.matrix.mvp.row[2], vertex.position;\n"
-      "DP4 vertexClip.w, state.matrix.mvp.row[3], vertex.position;\n"
-      "MOV result.position, vertexClip;\n"
-      "MOV result.texcoord[0], vertex.texcoord[0];\n"
-      "MOV result.color, vertex.color;\n"
-      "END\n";
-
-   assert(intel->meta.saved_vp == NULL);
-
-   _mesa_reference_vertprog(ctx, &intel->meta.saved_vp,
-			    ctx->VertexProgram.Current);
-   if (intel->meta.passthrough_vp == NULL) {
-      GLuint prog_name;
-      _mesa_GenPrograms(1, &prog_name);
-      _mesa_BindProgram(GL_VERTEX_PROGRAM_ARB, prog_name);
-      _mesa_ProgramStringARB(GL_VERTEX_PROGRAM_ARB,
-			     GL_PROGRAM_FORMAT_ASCII_ARB,
-			     strlen(vp), (const GLubyte *)vp);
-      _mesa_reference_vertprog(ctx, &intel->meta.passthrough_vp,
-			       ctx->VertexProgram.Current);
-      _mesa_DeletePrograms(1, &prog_name);
-   }
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
-			    intel->meta.passthrough_vp);
-   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
-			   &intel->meta.passthrough_vp->Base);
-
-   intel->meta.saved_vp_enable = ctx->VertexProgram.Enabled;
-   _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
-}
-
-/**
- * Restores the previous vertex program after
- * intel_meta_set_passthrough_vertex_program()
- */
-void
-intel_meta_restore_vertex_program(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
-			    intel->meta.saved_vp);
-   _mesa_reference_vertprog(ctx, &intel->meta.saved_vp, NULL);
-   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
-			   &ctx->VertexProgram.Current->Base);
-
-   if (!intel->meta.saved_vp_enable)
-      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
-}
-
-/**
- * Binds the given program string to GL_FRAGMENT_PROGRAM_ARB, caching the
- * program object.
- */
-void
-intel_meta_set_fragment_program(struct intel_context *intel,
-				struct gl_fragment_program **prog,
-				const char *prog_string)
-{
-   GLcontext *ctx = &intel->ctx;
-   assert(intel->meta.saved_fp == NULL);
-
-   _mesa_reference_fragprog(ctx, &intel->meta.saved_fp,
-			    ctx->FragmentProgram.Current);
-   if (*prog == NULL) {
-      GLuint prog_name;
-      _mesa_GenPrograms(1, &prog_name);
-      _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, prog_name);
-      _mesa_ProgramStringARB(GL_FRAGMENT_PROGRAM_ARB,
-			     GL_PROGRAM_FORMAT_ASCII_ARB,
-			     strlen(prog_string), (const GLubyte *)prog_string);
-      _mesa_reference_fragprog(ctx, prog, ctx->FragmentProgram.Current);
-      /* Note that DeletePrograms unbinds the program on us */
-      _mesa_DeletePrograms(1, &prog_name);
-   }
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, *prog);
-   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, &((*prog)->Base));
-
-   intel->meta.saved_fp_enable = ctx->FragmentProgram.Enabled;
-   _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
-}
-
-/**
- * Restores the previous fragment program after
- * intel_meta_set_fragment_program()
- */
-void
-intel_meta_restore_fragment_program(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current,
-			    intel->meta.saved_fp);
-   _mesa_reference_fragprog(ctx, &intel->meta.saved_fp, NULL);
-   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB,
-			   &ctx->FragmentProgram.Current->Base);
-
-   if (!intel->meta.saved_fp_enable)
-      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
-}
-
-static const float default_texcoords[4][2] = { { 0.0, 0.0 },
-					       { 1.0, 0.0 },
-					       { 1.0, 1.0 },
-					       { 0.0, 1.0 } };
-
-void
-intel_meta_set_default_texrect(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-   struct gl_client_array *old_texcoord_array;
-
-   intel->meta.saved_active_texture = ctx->Texture.CurrentUnit;
-   if (intel->meta.saved_array_vbo == NULL) {
-      _mesa_reference_buffer_object(ctx, &intel->meta.saved_array_vbo,
-				    ctx->Array.ArrayBufferObj);
-   }
-
-   old_texcoord_array = &ctx->Array.ArrayObj->TexCoord[0];
-   intel->meta.saved_texcoord_type = old_texcoord_array->Type;
-   intel->meta.saved_texcoord_size = old_texcoord_array->Size;
-   intel->meta.saved_texcoord_stride = old_texcoord_array->Stride;
-   intel->meta.saved_texcoord_enable = old_texcoord_array->Enabled;
-   intel->meta.saved_texcoord_ptr = old_texcoord_array->Ptr;
-   _mesa_reference_buffer_object(ctx, &intel->meta.saved_texcoord_vbo,
-				 old_texcoord_array->BufferObj);
-
-   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
-
-   if (intel->meta.texcoord_vbo == NULL) {
-      GLuint vbo_name;
-
-      _mesa_GenBuffersARB(1, &vbo_name);
-      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, vbo_name);
-      _mesa_BufferDataARB(GL_ARRAY_BUFFER_ARB, sizeof(default_texcoords),
-			  default_texcoords, GL_STATIC_DRAW_ARB);
-      _mesa_reference_buffer_object(ctx, &intel->meta.texcoord_vbo,
-				    ctx->Array.ArrayBufferObj);
-   } else {
-      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
-			  intel->meta.texcoord_vbo->Name);
-   }
-   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), NULL);
-
-   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
-}
-
-void
-intel_meta_restore_texcoords(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   /* Restore the old TexCoordPointer */
-   if (intel->meta.saved_texcoord_vbo) {
-      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
-			  intel->meta.saved_texcoord_vbo->Name);
-      _mesa_reference_buffer_object(ctx, &intel->meta.saved_texcoord_vbo, NULL);
-   } else {
-      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
-   }
-
-   _mesa_TexCoordPointer(intel->meta.saved_texcoord_size,
-			 intel->meta.saved_texcoord_type,
-			 intel->meta.saved_texcoord_stride,
-			 intel->meta.saved_texcoord_ptr);
-   if (!intel->meta.saved_texcoord_enable)
-      _mesa_Disable(GL_TEXTURE_COORD_ARRAY);
-
-   _mesa_ClientActiveTextureARB(GL_TEXTURE0 +
-				intel->meta.saved_active_texture);
-
-   if (intel->meta.saved_array_vbo) {
-      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
-			  intel->meta.saved_array_vbo->Name);
-      _mesa_reference_buffer_object(ctx, &intel->meta.saved_array_vbo, NULL);
-   } else {
-      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
-   }
-}
-
 void
 intelInitPixelFuncs(struct dd_function_table *functions)
 {
@@ -428,14 +188,3 @@ intelInitPixelFuncs(struct dd_function_table *functions)
    functions->ReadPixels = intelReadPixels;
 }
 
-void
-intel_free_pixel_state(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   _mesa_reference_vertprog(ctx, &intel->meta.passthrough_vp, NULL);
-   _mesa_reference_fragprog(ctx, &intel->meta.bitmap_fp, NULL);
-   _mesa_reference_fragprog(ctx, &intel->meta.tex2d_fp, NULL);
-   _mesa_reference_buffer_object(ctx, &intel->meta.texcoord_vbo, NULL);
-}
-
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.h b/src/mesa/drivers/dri/intel/intel_pixel.h
index 6acf0813c8..96a6dd17b2 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.h
+++ b/src/mesa/drivers/dri/intel/intel_pixel.h
@@ -31,19 +31,6 @@
 #include "main/mtypes.h"
 
 void intelInitPixelFuncs(struct dd_function_table *functions);
-void intel_meta_set_passthrough_transform(struct intel_context *intel);
-void intel_meta_restore_transform(struct intel_context *intel);
-void intel_meta_set_passthrough_vertex_program(struct intel_context *intel);
-void intel_meta_restore_vertex_program(struct intel_context *intel);
-void intel_meta_set_fragment_program(struct intel_context *intel,
-				     struct gl_fragment_program **prog,
-				     const char *prog_string);
-void intel_meta_restore_fragment_program(struct intel_context *intel);
-void intel_free_pixel_state(struct intel_context *intel);
-void intel_meta_set_default_texrect(struct intel_context *intel);
-void intel_meta_set_default_texrect(struct intel_context *intel);
-void intel_meta_restore_texcoords(struct intel_context *intel);
-
 GLboolean intel_check_blit_fragment_ops(GLcontext * ctx,
 					GLboolean src_alpha_is_one);
 
@@ -79,6 +66,4 @@ void intelBitmap(GLcontext * ctx,
 		 const struct gl_pixelstore_attrib *unpack,
 		 const GLubyte * pixels);
 
-void intel_clear_tris(GLcontext *ctx, GLbitfield mask);
-
 #endif
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
index ebba6aafad..540e7620a9 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
@@ -476,11 +476,11 @@ intel_texture_bitmap(GLcontext * ctx,
 		    GL_ALPHA, GL_UNSIGNED_BYTE, a8_bitmap);
    _mesa_free(a8_bitmap);
 
-   intel_meta_set_fragment_program(intel, &intel->meta.bitmap_fp, fp);
+   meta_set_fragment_program(&intel->meta, &intel->meta.bitmap_fp, fp);
    _mesa_ProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0,
 				     ctx->Current.RasterColor);
-   intel_meta_set_passthrough_vertex_program(intel);
-   intel_meta_set_passthrough_transform(intel);
+   meta_set_passthrough_vertex_program(&intel->meta);
+   meta_set_passthrough_transform(&intel->meta);
 
    /* convert rasterpos Z from [0,1] to NDC coord in [-1,1] */
    dst_z = -1.0 + 2.0 * ctx->Current.RasterPos[2];
@@ -507,13 +507,13 @@ intel_texture_bitmap(GLcontext * ctx,
 
    _mesa_VertexPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &vertices);
    _mesa_Enable(GL_VERTEX_ARRAY);
-   intel_meta_set_default_texrect(intel);
+   meta_set_default_texrect(&intel->meta);
    _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
 
-   intel_meta_restore_texcoords(intel);
-   intel_meta_restore_transform(intel);
-   intel_meta_restore_fragment_program(intel);
-   intel_meta_restore_vertex_program(intel);
+   meta_restore_texcoords(&intel->meta);
+   meta_restore_transform(&intel->meta);
+   meta_restore_fragment_program(&intel->meta);
+   meta_restore_vertex_program(&intel->meta);
 
    _mesa_PopClientAttrib();
    _mesa_Disable(GL_TEXTURE_2D); /* asserted that it was disabled at entry */
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_draw.c b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
index dfc9e1539f..a6b6824164 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_draw.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
@@ -150,7 +150,7 @@ intel_texture_drawpixels(GLcontext * ctx,
    _mesa_TexImage2D(GL_TEXTURE_2D, 0, internalFormat, width, height, 0, format,
 		    type, pixels);
 
-   intel_meta_set_passthrough_transform(intel);
+   meta_set_passthrough_transform(&intel->meta);
 
    /* convert rasterpos Z from [0,1] to NDC coord in [-1,1] */
    z = -1.0 + 2.0 * ctx->Current.RasterPos[2];
@@ -182,12 +182,12 @@ intel_texture_drawpixels(GLcontext * ctx,
 
    _mesa_VertexPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &vertices);
    _mesa_Enable(GL_VERTEX_ARRAY);
-   intel_meta_set_default_texrect(intel);
+   meta_set_default_texrect(&intel->meta);
 
    _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
 
-   intel_meta_restore_texcoords(intel);
-   intel_meta_restore_transform(intel);
+   meta_restore_texcoords(&intel->meta);
+   meta_restore_transform(&intel->meta);
 
    _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
    _mesa_PopClientAttrib();
@@ -352,7 +352,7 @@ intel_stencil_drawpixels(GLcontext * ctx,
    ctx->Unpack = old_unpack;
    _mesa_free(stencil_pixels);
 
-   intel_meta_set_passthrough_transform(intel);
+   meta_set_passthrough_transform(&intel->meta);
 
    /* Since we're rendering to the framebuffer as if it was an FBO,
     * if it's the window system we have to flip the coordinates.
@@ -375,12 +375,12 @@ intel_stencil_drawpixels(GLcontext * ctx,
 
    _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices);
    _mesa_Enable(GL_VERTEX_ARRAY);
-   intel_meta_set_default_texrect(intel);
+   meta_set_default_texrect(&intel->meta);
 
    _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
 
-   intel_meta_restore_texcoords(intel);
-   intel_meta_restore_transform(intel);
+   meta_restore_texcoords(&intel->meta);
+   meta_restore_transform(&intel->meta);
 
    _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
    _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, old_fb_name);
diff --git a/src/mesa/drivers/dri/r200/r200_ioctl.c b/src/mesa/drivers/dri/r200/r200_ioctl.c
index 6560efdca3..4dbda39eb9 100644
--- a/src/mesa/drivers/dri/r200/r200_ioctl.c
+++ b/src/mesa/drivers/dri/r200/r200_ioctl.c
@@ -58,11 +58,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define R200_TIMEOUT             512
 #define R200_IDLE_RETRY           16
 
-static void r200UserClear(GLcontext *ctx, GLuint mask)
-{
-   radeon_clear_tris(ctx, mask);
-}
-
 static void r200KernelClear(GLcontext *ctx, GLuint flags)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
@@ -253,7 +248,7 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask )
    }
 
    if (rmesa->radeon.radeonScreen->kernel_mm)
-      r200UserClear(ctx, orig_mask);
+      radeonUserClear(ctx, orig_mask);
    else {
       r200KernelClear(ctx, flags);
       rmesa->radeon.hw.all_dirty = GL_TRUE;
diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.c b/src/mesa/drivers/dri/r300/r300_ioctl.c
index 4e913dba29..ddabd53992 100644
--- a/src/mesa/drivers/dri/r300/r300_ioctl.c
+++ b/src/mesa/drivers/dri/r300/r300_ioctl.c
@@ -68,11 +68,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 static void r300EmitClearState(GLcontext * ctx);
 
-static void r300UserClear(GLcontext *ctx, GLuint mask)
-{
-	radeon_clear_tris(ctx, mask);
-}
-
 static void r300ClearBuffer(r300ContextPtr r300, int flags,
 			    struct radeon_renderbuffer *rrb,
 			    struct radeon_renderbuffer *rrbd)
@@ -680,7 +675,7 @@ static void r300Clear(GLcontext * ctx, GLbitfield mask)
 	ret = 0;
 	if (tri_mask) {
 		if (r300->radeon.radeonScreen->kernel_mm)
-			r300UserClear(ctx, tri_mask);
+			radeonUserClear(ctx, tri_mask);
 		else {
 			/* if kernel clear fails due to size restraints fallback */
 			ret = r300KernelClear(ctx, tri_mask);
diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c
index 0a8d8b03e8..3bf42e9bb0 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common.c
+++ b/src/mesa/drivers/dri/radeon/radeon_common.c
@@ -877,7 +877,7 @@ void radeon_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei he
 	if (!driContext->driScreenPriv->dri2.enabled)
 		return;
 
-	if (!radeon->internal_viewport_call && ctx->DrawBuffer->Name == 0) {
+	if (!radeon->meta.internal_viewport_call && ctx->DrawBuffer->Name == 0) {
 		if (radeon->is_front_buffer_rendering) {
 			radeonFlush(ctx);
 		}
@@ -1258,254 +1258,8 @@ void rcommonBeginBatch(radeonContextPtr rmesa, int n,
 
 }
 
-
-
-static void
-radeon_meta_set_passthrough_transform(radeonContextPtr radeon)
-{
-   GLcontext *ctx = radeon->glCtx;
-
-   radeon->meta.saved_vp_x = ctx->Viewport.X;
-   radeon->meta.saved_vp_y = ctx->Viewport.Y;
-   radeon->meta.saved_vp_width = ctx->Viewport.Width;
-   radeon->meta.saved_vp_height = ctx->Viewport.Height;
-   radeon->meta.saved_matrix_mode = ctx->Transform.MatrixMode;
-
-   radeon->internal_viewport_call = GL_TRUE;
-   _mesa_Viewport(0, 0, ctx->DrawBuffer->Width, ctx->DrawBuffer->Height);
-   radeon->internal_viewport_call = GL_FALSE;
-
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-   _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1);
-
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-}
-
-static void
-radeon_meta_restore_transform(radeonContextPtr radeon)
-{
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PopMatrix();
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PopMatrix();
-
-   _mesa_MatrixMode(radeon->meta.saved_matrix_mode);
-
-   radeon->internal_viewport_call = GL_TRUE;
-   _mesa_Viewport(radeon->meta.saved_vp_x, radeon->meta.saved_vp_y,
-		  radeon->meta.saved_vp_width, radeon->meta.saved_vp_height);
-   radeon->internal_viewport_call = GL_FALSE;
-}
-
-
-/**
- * Perform glClear where mask contains only color, depth, and/or stencil.
- *
- * The implementation is based on calling into Mesa to set GL state and
- * performing normal triangle rendering.  The intent of this path is to
- * have as generic a path as possible, so that any driver could make use of
- * it.
- */
-
-static void radeon_clear_init(GLcontext *ctx)
-{
-    radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-    struct gl_array_object *arraySave = NULL;
-    const GLuint arrayBuffer = ctx->Array.ArrayBufferObj->Name;
-    const GLuint elementBuffer = ctx->Array.ElementArrayBufferObj->Name;
-
-    /* create new array object */
-    rmesa->clear.arrayObj = _mesa_new_array_object(ctx, ~0);
-    _mesa_reference_array_object(ctx, &arraySave, ctx->Array.ArrayObj);
-    _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, rmesa->clear.arrayObj);
-
-    /* one time setup of vertex arrays (pos, color) */
-    _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
-    _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0);
-    _mesa_ColorPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), rmesa->clear.color);
-    _mesa_VertexPointer(3, GL_FLOAT, 3 * sizeof(GLfloat), rmesa->clear.vertices);
-    _mesa_Enable(GL_COLOR_ARRAY);
-    _mesa_Enable(GL_VERTEX_ARRAY);
-
-    /* restore original array object */
-    _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, arraySave);
-    _mesa_reference_array_object(ctx, &arraySave, NULL);
-
-    /* restore original buffer objects */
-    _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, arrayBuffer);
-    _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, elementBuffer);
-}
-
-
-void radeon_clear_tris(GLcontext *ctx, GLbitfield mask)
+void radeonUserClear(GLcontext *ctx, GLuint mask)
 {
    radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat dst_z;
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   int i;
-   GLboolean saved_fp_enable = GL_FALSE, saved_vp_enable = GL_FALSE;
-   GLboolean saved_shader_program = 0;
-   unsigned int saved_active_texture;
-   struct gl_array_object *arraySave = NULL;
-
-   if (!rmesa->clear.arrayObj)
-       radeon_clear_init(ctx);
-
-   assert((mask & ~(TRI_CLEAR_COLOR_BITS | BUFFER_BIT_DEPTH |
-		    BUFFER_BIT_STENCIL)) == 0);
-
-   _mesa_PushAttrib(GL_COLOR_BUFFER_BIT |
-		    GL_DEPTH_BUFFER_BIT |
-		    GL_ENABLE_BIT |
-		    GL_POLYGON_BIT |
-		    GL_STENCIL_BUFFER_BIT |
-		    GL_TRANSFORM_BIT |
-		    GL_CURRENT_BIT);
-   saved_active_texture = ctx->Texture.CurrentUnit;
-
-  /* Disable existing GL state we don't want to apply to a clear. */
-   _mesa_Disable(GL_ALPHA_TEST);
-   _mesa_Disable(GL_BLEND);
-   _mesa_Disable(GL_CULL_FACE);
-   _mesa_Disable(GL_FOG);
-   _mesa_Disable(GL_POLYGON_SMOOTH);
-   _mesa_Disable(GL_POLYGON_STIPPLE);
-   _mesa_Disable(GL_POLYGON_OFFSET_FILL);
-   _mesa_Disable(GL_LIGHTING);
-   _mesa_Disable(GL_CLIP_PLANE0);
-   _mesa_Disable(GL_CLIP_PLANE1);
-   _mesa_Disable(GL_CLIP_PLANE2);
-   _mesa_Disable(GL_CLIP_PLANE3);
-   _mesa_Disable(GL_CLIP_PLANE4);
-   _mesa_Disable(GL_CLIP_PLANE5);
-   _mesa_PolygonMode(GL_FRONT_AND_BACK, GL_FILL);
-   if (ctx->Extensions.ARB_fragment_program && ctx->FragmentProgram.Enabled) {
-      saved_fp_enable = GL_TRUE;
-      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
-   }
-   if (ctx->Extensions.ARB_vertex_program && ctx->VertexProgram.Enabled) {
-      saved_vp_enable = GL_TRUE;
-      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
-   }
-   if (ctx->Extensions.ARB_shader_objects && ctx->Shader.CurrentProgram) {
-      saved_shader_program = ctx->Shader.CurrentProgram->Name;
-      _mesa_UseProgramObjectARB(0);
-   }
-
-   if (ctx->Texture._EnabledUnits != 0) {
-      int i;
-
-      for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-	 _mesa_ActiveTextureARB(GL_TEXTURE0 + i);
-	 _mesa_Disable(GL_TEXTURE_1D);
-	 _mesa_Disable(GL_TEXTURE_2D);
-	 _mesa_Disable(GL_TEXTURE_3D);
-	 if (ctx->Extensions.ARB_texture_cube_map)
-	    _mesa_Disable(GL_TEXTURE_CUBE_MAP_ARB);
-	 if (ctx->Extensions.NV_texture_rectangle)
-	    _mesa_Disable(GL_TEXTURE_RECTANGLE_NV);
-	 if (ctx->Extensions.MESA_texture_array) {
-	    _mesa_Disable(GL_TEXTURE_1D_ARRAY_EXT);
-	    _mesa_Disable(GL_TEXTURE_2D_ARRAY_EXT);
-	 }
-      }
-   }
-
-   /* save current array object, bind our private one */
-   _mesa_reference_array_object(ctx, &arraySave, ctx->Array.ArrayObj);
-   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, rmesa->clear.arrayObj);
-
-   radeon_meta_set_passthrough_transform(rmesa);
-
-   for (i = 0; i < 4; i++) {
-      COPY_4FV(rmesa->clear.color[i], ctx->Color.ClearColor);
-   }
-
-   /* convert clear Z from [0,1] to NDC coord in [-1,1] */
-
-   dst_z = -1.0 + 2.0 * ctx->Depth.Clear;
-   /* Prepare the vertices, which are the same regardless of which buffer we're
-    * drawing to.
-    */
-   rmesa->clear.vertices[0][0] = fb->_Xmin;
-   rmesa->clear.vertices[0][1] = fb->_Ymin;
-   rmesa->clear.vertices[0][2] = dst_z;
-   rmesa->clear.vertices[1][0] = fb->_Xmax;
-   rmesa->clear.vertices[1][1] = fb->_Ymin;
-   rmesa->clear.vertices[1][2] = dst_z;
-   rmesa->clear.vertices[2][0] = fb->_Xmax;
-   rmesa->clear.vertices[2][1] = fb->_Ymax;
-   rmesa->clear.vertices[2][2] = dst_z;
-   rmesa->clear.vertices[3][0] = fb->_Xmin;
-   rmesa->clear.vertices[3][1] = fb->_Ymax;
-   rmesa->clear.vertices[3][2] = dst_z;
-
-   while (mask != 0) {
-      GLuint this_mask = 0;
-      GLuint color_bit;
-
-      color_bit = _mesa_ffs(mask & TRI_CLEAR_COLOR_BITS);
-      if (color_bit != 0)
-	 this_mask |= (1 << (color_bit - 1));
-
-      /* Clear depth/stencil in the same pass as color. */
-      this_mask |= (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL));
-
-      /* Select the current color buffer and use the color write mask if
-       * we have one, otherwise don't write any color channels.
-       */
-      if (this_mask & BUFFER_BIT_FRONT_LEFT)
-	 _mesa_DrawBuffer(GL_FRONT_LEFT);
-      else if (this_mask & BUFFER_BIT_BACK_LEFT)
-	 _mesa_DrawBuffer(GL_BACK_LEFT);
-      else if (color_bit != 0)
-	 _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0 +
-			  (color_bit - BUFFER_COLOR0 - 1));
-      else
-	 _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
-
-      /* Control writing of the depth clear value to depth. */
-      if (this_mask & BUFFER_BIT_DEPTH) {
-	 _mesa_DepthFunc(GL_ALWAYS);
-	 _mesa_DepthMask(GL_TRUE);
-	 _mesa_Enable(GL_DEPTH_TEST);
-      } else {
-	 _mesa_Disable(GL_DEPTH_TEST);
-	 _mesa_DepthMask(GL_FALSE);
-      }
-
-      /* Control writing of the stencil clear value to stencil. */
-      if (this_mask & BUFFER_BIT_STENCIL) {
-	 _mesa_Enable(GL_STENCIL_TEST);
-	 _mesa_StencilOp(GL_REPLACE, GL_REPLACE, GL_REPLACE);
-	 _mesa_StencilFuncSeparate(GL_FRONT_AND_BACK, GL_ALWAYS, (GLint)ctx->Stencil.Clear,
-				   ctx->Stencil.WriteMask[0]);
-      } else {
-	 _mesa_Disable(GL_STENCIL_TEST);
-      }
-
-      _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
-
-      mask &= ~this_mask;
-   }
-
-   radeon_meta_restore_transform(rmesa);
-
-   _mesa_ActiveTextureARB(GL_TEXTURE0 + saved_active_texture);
-   if (saved_fp_enable)
-      _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
-   if (saved_vp_enable)
-      _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
-
-   if (saved_shader_program)
-      _mesa_UseProgramObjectARB(saved_shader_program);
-
-   _mesa_PopAttrib();
-     /* restore current array object */
-   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, arraySave);
-   _mesa_reference_array_object(ctx, &arraySave, NULL);
+   meta_clear_tris(&rmesa->meta, mask);
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_common.h b/src/mesa/drivers/dri/radeon/radeon_common.h
index ba6c7c5773..cebae18b2d 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common.h
+++ b/src/mesa/drivers/dri/radeon/radeon_common.h
@@ -5,18 +5,7 @@
 #include "radeon_dma.h"
 #include "radeon_texture.h"
 
-
-#define TRI_CLEAR_COLOR_BITS (BUFFER_BIT_BACK_LEFT |			\
-			      BUFFER_BIT_FRONT_LEFT |			\
-			      BUFFER_BIT_COLOR0 |			\
-			      BUFFER_BIT_COLOR1 |			\
-			      BUFFER_BIT_COLOR2 |			\
-			      BUFFER_BIT_COLOR3 |			\
-			      BUFFER_BIT_COLOR4 |			\
-			      BUFFER_BIT_COLOR5 |			\
-			      BUFFER_BIT_COLOR6 |			\
-			      BUFFER_BIT_COLOR7)
-
+void radeonUserClear(GLcontext *ctx, GLuint mask);
 void radeonRecalcScissorRects(radeonContextPtr radeon);
 void radeonSetCliprects(radeonContextPtr radeon);
 void radeonUpdateScissor( GLcontext *ctx );
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c
index 828d6477f0..9add50b4cc 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c
@@ -184,6 +184,7 @@ GLboolean radeonInitContext(radeonContextPtr radeon,
 	ctx = radeon->glCtx;
 	driContextPriv->driverPrivate = radeon;
 
+	meta_init_metaops(ctx, &radeon->meta);
 	/* DRI fields */
 	radeon->dri.context = driContextPriv;
 	radeon->dri.screen = sPriv;
@@ -264,7 +265,7 @@ void radeonDestroyContext(__DRIcontextPrivate *driContextPriv )
 		}
 
 		radeonReleaseArrays(radeon->glCtx, ~0);
-
+		meta_destroy_metaops(&radeon->meta);
 		if (radeon->vtbl.free_context)
 			radeon->vtbl.free_context(radeon->glCtx);
 		_swsetup_DestroyContext( radeon->glCtx );
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h
index 036d2658d9..07ac85fb52 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common_context.h
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h
@@ -13,6 +13,7 @@
 #include "dri_util.h"
 #include "tnl/t_vertex.h"
 
+#include "dri_metaops.h"
 struct radeon_context;
 
 #include "radeon_bocs_wrapper.h"
@@ -476,27 +477,7 @@ struct radeon_context {
     */
    GLboolean is_front_buffer_reading;
 
-   /* info for radeon_clear_tris() */
-   struct {
-      struct gl_array_object *arrayObj;
-      GLfloat vertices[4][3];
-      GLfloat color[4][4];
-   } clear;
-   GLboolean internal_viewport_call;
-
-  struct {
-      struct gl_fragment_program *bitmap_fp;
-      struct gl_vertex_program *passthrough_vp;
-
-      struct gl_fragment_program *saved_fp;
-      GLboolean saved_fp_enable;
-      struct gl_vertex_program *saved_vp;
-      GLboolean saved_vp_enable;
-
-      GLint saved_vp_x, saved_vp_y;
-      GLsizei saved_vp_width, saved_vp_height;
-      GLenum saved_matrix_mode;
-   } meta;
+   struct dri_metaops meta;
 
    struct {
 	   void (*get_lock)(radeonContextPtr radeon);
diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c
index 01c45df2df..a5e4df7941 100644
--- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c
@@ -445,11 +445,6 @@ void radeonEmitAOS( r100ContextPtr rmesa,
  */
 #define RADEON_MAX_CLEARS	256
 
-static void radeonUserClear(GLcontext *ctx, GLuint mask)
-{
-   radeon_clear_tris(ctx, mask);
-}
-
 static void radeonKernelClear(GLcontext *ctx, GLuint flags)
 {
      r100ContextPtr rmesa = R100_CONTEXT(ctx);
-- 
cgit v1.2.3


From bd4ed25222987f80b720325204e300fd79f573cf Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Wed, 15 Jul 2009 13:24:30 +1000
Subject: r300: emit z depth pitch reloc in preparation for tiling

---
 src/mesa/drivers/dri/r300/r300_cmdbuf.c | 24 +++++++++++++++++-------
 1 file changed, 17 insertions(+), 7 deletions(-)

(limited to 'src/mesa/drivers/dri/r300')

diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
index 248297897d..af535037d0 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
@@ -344,23 +344,33 @@ static void emit_zb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
 	BATCH_LOCALS(&r300->radeon);
 	struct radeon_renderbuffer *rrb;
 	uint32_t zbpitch;
+	uint32_t dw;
 
 	rrb = radeon_get_depthbuffer(&r300->radeon);
 	if (!rrb)
 		return;
 
 	zbpitch = (rrb->pitch / rrb->cpp);
-	if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) {
-		zbpitch |= R300_DEPTHMACROTILE_ENABLE;
-	}
-	if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE){
-		zbpitch |= R300_DEPTHMICROTILE_TILED;
+	if (!r300->radeon.radeonScreen->kernel_mm) {
+	    if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) {
+	        zbpitch |= R300_DEPTHMACROTILE_ENABLE;
+	   }
+	    if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE){
+	        zbpitch |= R300_DEPTHMICROTILE_TILED;
+	    }
 	}
 
-	BEGIN_BATCH_NO_AUTOSTATE(6);
+	dw = 6;
+    	if (r300->radeon.radeonScreen->kernel_mm)
+		dw += 2;
+	BEGIN_BATCH_NO_AUTOSTATE(dw);
 	OUT_BATCH_REGSEQ(R300_ZB_DEPTHOFFSET, 1);
 	OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
-	OUT_BATCH_REGVAL(R300_ZB_DEPTHPITCH, zbpitch);
+	OUT_BATCH_REGSEQ(R300_ZB_DEPTHPITCH, 1);
+    	if (!r300->radeon.radeonScreen->kernel_mm)
+	    OUT_BATCH(zbpitch);
+	else
+	    OUT_BATCH_RELOC(cbpitch, rrb->bo, zbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0);
 	END_BATCH();
 }
 
-- 
cgit v1.2.3